Multi-Channel Attribution: Definitions, Models and a Reality Check

Yum A wise person said: "To guarantee success, spend 95% of your time defining the problem and 5% of the time solving it."

I believe deeply in that quote. In my life I spend an extraordinary amount of time understanding the problem and attempting to define it clearly. As if by magic, I find that it is then much easier to find the optimal solution (or realize none exists!).

Multi-Channel Attribution is a red hot topic in our industry, and yet it is so poorly understood. I'm convinced that the resulting problems (confusion, FUD, angst, daily prayers, and wasted budget) are due to the lack of a clear framework that can help clearly define the problem.

In this post my hope is share a framework that will help define the problem clearly. Included in the post are recommendations for measurement and data analysis. And as if that was not enough, :), I'll close the post with my thoughts on digital marketing attribution models.

This is going to be a lot of fun. Roll up your sleeves, put a smile on your face, grab a pinch of common sense, a heavy dose of reality and let's go…

Three Types of Multi-Channel Attribution Problems.

A huge amount of confusion and disagreement on this topic exists simply because there is no general consensus about those three words. Multi-Channel Attribution.

So let's try and fix that problem.

There are three types of attribution problems in our non-line world.

Multi-Channel Attribution, Online to Store:

This is the attempt by Marketers and Analysts to try and understand the offline impact (revenue/brand value/butts in seats/phone calls/etc) driven by online marketing and advertising. We'll refer to this quest for doing effective attribution as MCA-O2S.

While I'm using the term Store here, it encompasses sales (or leads or catalog requests) driven to a retail store or company call center, people driven to donate blood via online campaigns, or essentially any offline outcome driven by the online channel.

An example of MCA-O2S is Verizon wanting to know how many in-store offline phone activations are driven by online search advertising, for every online activation that the same search advertising drives.

[In case you were curious… It's 5 new accounts activated offline for every 1 activated online! If you are not calculating the offline impact, and you are not giving your online channel due credit. In this case, it would be 5x less credit! You can see why MCA-O2S is supremely critical for every company on the planet. You can also download the slides from a HP online-to-store experiment and how they learned that every $1 spent on online advertising delivered $5.3 in offline store sales. Or watch a video of how Quaker Oats boosted sales by 9% in store via YouTube videos.]

Here's the Post-It on which I'd sketched MCA-O2S in planning this post. The red dots represent activity we would like to ensure we are measuring to 1) ensure we understand behavior, and 2) deliver insights that will influence our marketing and advertising…

Multi-Channel Attribution Online to Store

I spend a lot of time with CEOs and CMOs and when they talk about multi-channel attribution, they're invariably talking about MCA-O2S. Yet when most of my digital peers talk about this topic, they're not talking about MCA-O2S. You can imagine why things might get a little confusing.

So when you meet a CEO and they use say "Help me solve the amazing multi-channel attribution problem", you say: "which type of MCA are you interested in?" Clarity will help foster a valuable conversation.

Almost all current, hot and heavy, literature on the topic of attribution modeling does not cover MCA-O2S. That's because when it comes to MCA-O2S your only bffs are a set of 16 strategies I've outlined over two posts (links immediately below) or the fantastic world of controlled experiments (as in the Verizon case above). So less automated algorithms "distributing credit" and more thoughtful deliberative discreet measurement strategies that inform strategic decisions.

Two helpful blog posts on multi-channel analytics: 1. Tracking online impact of offline advertising. 2. Tracking offline impact of online advertising.

MCA-O2S. It's mandatory. Attribution is driven by experiments. And when you win, you win huge!

Multi-Channel Attribution, Across Multiple Screens:

Senior leaders, especially in larger companies, have started to refer to this when they use the magical words multi-channel attribution.

With the massive adoption of mobile phones and tablets we are all increasingly "four screen" people (TV, desktop, tablets, smart phones). That has directly translated into a more complex fragmented influence landscape (drives the "old timers" bananas). That in turn has translated into many senior leaders deeply desiring, as they put it, "multi-channel attribution." What they really mean is MCA-AMS.

What they really really want is to understand how individuals experience a company or government's digital existence across multiple devices, what media (advertising and marketing) they are being exposed to, and what outcomes (conversions!) are happening as a result.

An example of MCA-AMS is the ability to understand that a search I did on my tablet computer while watching a television commercial resulted in a click on a paid search ad to a camera site which logged into my memory which later caused me to read reviews of the camera on my Nexus S while stuck in traffic and that finally caused a sale for Sony when I got home and happened to be on my laptop.

Attribution in this case is the quest to apportion credit across the TV commercial, tablet paid search ad, reviews read on on the mobile phone for a "direct" conversion on the PC. Amazing, right?

Here's my sketch on MCA-AMS and the raw complexity of the customer experience that we are trying to understand… the red dots indicate what we're trying to measure and understand the impact of…

Multi-Channel Attribution Across Multiple Screens

The primary challenge is that as we switch devices it is increasingly difficult to keep track of the same person as they interface with our digital existence (and are exposed to online and offline marketing and advertising). Actually, I should not say increasingly difficult, I should say almost impossible (cookies, uuids, privacy, government, et al).

Perhaps the only exceptions to the "its almost impossible" scenario would be companies that service customers who are mostly logged in (think Amazon, NY Times) across all four screens all the time. Such companies usually also own massive data warehouses where they have an ability to periodically do cannonballs into the data and identify correlations in consumption and purchase patterns. Often, though not always, they can also tease out causations between devices used during outcomes (five-second segmentation in say Google Analytics) and their media plans while focusing on customer analysis (not visitors, not cookies, not uuids, customers).

Even then it is hard, very hard. And for the rest of us this will remain a complex, and I'm sorry to be so real, unsolvable challenge. At least for now.

Some ideas from the two multi-channel blog posts above can help with MCA-AMS. I've leveraged controlled experiments to get very good "kinda sorta understanding" of reality.

I believe that real solutions will come from the evolution of cookies, updating privacy policies, government decisions and evolving user habits. All that first, then our ability to have nonline data.

Because of all of the above you can see why attribution models don't even enter the picture with MCA-AMS. But when you meet executives and they say "help us with our multi-channel attribution problem", most definitely ask the clarifying question: "do you mean MCA-O2S or MCA-AMS?"

MCA-AMS. Complex, hard challenge. Not a huge problem yet for most, but heading in that direction.

Multi-Channel Attribution, Across Digital Channels:

Almost all of the time when people in our ecosystem (unlike CEOs, CMOs) talk about Multi-Channel Attribution, this is the one they are referring to.

MCA-ADC is the effort to understand which digital marketing channels (Social, Display, YouTube, Referral, Email, Search, others) contributed to a particular conversion (or multiple conversions).

At the moment all web analytics tools, like SiteCatalyst, WebTrends, Google Analytics, CoreMetrics, and others, by default attribute a conversion to the channel immediately prior to the conversion. This is also known as last click attribution.

With MCA-ADC you are trying to go beyond the last click and get this, complete, picture of all marketing activity prior to the conversion (in this case from Google Analytics):

digital marketing path to conversion

For this website, 767 conversions came from people who visited the site in the above precise order starting with social then a direct visit then an organic search then a referral click-through and finally one last direct visit which lead to the conversions.

The attribution bit here is the burning desire inside all digital marketers to figure out how to dole out credit for the above conversions. Should Direct get 50%? How about Social? 100%? What about Organic? 2%? But let's put that delightful thought on the back burner for just a minute while we understand a critical, often hidden, nuance. [Analysis Ninjas are magnificent at understanding nuance!]

When people talk about MCA-ADC they are still just talking about one device. Because in very close to 0% of the cases do any of these analytics tool have an idea about the behavior of one homo sapien across multiple screens (AMS).

So what you are seeing above are all the conversions that can be tied to multiple visits by a unique browser (notice I did not say person) to your website/digital existence. BTW it is fantastic that GA does this because most other tools don't even show you this.

Say, the Organic Search above had happened on a mobile phone… regardless of the digital analytics tool used, to most websites today that visit would be invisible in the above chain (cookies!). #omg

Hence it is important to separate out MCA-AMS (across multiple screens) from MCA-ADC (across digital channels) – at least for now, until the cookies, ids, privacy policies, government guidance and user habits problem is solved.

When it comes to measuring MCA-AMS you'll use the guidance provided in the above section. For MCA-ADC you'll use a different set of reports (multi-channel funnels ) and attribution models.

I'm sure you are already familiar with nuance number two when it comes to MCA-ADC. A blind-spot if you will.

The above picture does not capture what the impact of this behavior was on your offline existence (O2S). Web Analytics tools are not awesome at that. Ok, they stink at it.

So it is possible that an additional 3,835 people went and made purchases in your stores or via your phone channel (taking the Verizon numbers from above). That would also be invisible from the above report. None of the channels above, whether glorious social, beloved direct, magnificent search, sweet referral, would ever get "credit." Unless you are willing to use the methodologies outlined in the MCA-O2S section above.

When you talk about MCA-ADC, ensure that you are aware and communicate to your leadership, that you are not reporting on MCA-O2S (online to store) and it is extremely unlikely to be reporting the impact of MCA-AMS.

Here's one last Post-It sketch. The red dots are what you are likely measuring when you attempt MCA-ADC…

Multi-Channel Attribution All Digital Channels

And if I wanted to be pedantic I would say it is really MCA-ADCFOD. Multi-channel attribution across digital channels for one device.

Now it is true that with sufficient analytical skills, time, patience, and God's direct blessing to you, it might be possible to do complete multi-channel attribution analysis where the multi-channel includes multiple online ad channels, behavior of the person across devices and the impact online and offline. Sadly, that is incredibly hard to do as a whole. And when I say incredibly hard, I mean almost impossible. And when I say almost impossible, I mean only attempt that after you know you've fixed all other problems with your advertising, your online and offline existence and your people. All three.

I know that sounds like a bummer, but a dose of reality is particularly needed in this discussion. There are simply too many fake promises being made by vendors, consultants, tweeters, gurus and fairies. That is unhelpful to the entire ecosystem.

To close this section…

Next time you hear someone utter the words multi-channel attribution, the single greatest gift you can give yourself is to ask in your sweetest possible voice: "Are you referring to MCA-O2S, MCA-AMS or MCA-ADC?"

You'll earn their respect for knowing that there are three types, and you'll be able to put into context what they are asking for and proceed to have a career and business-enhancing discussion.

Multi-Channel Attribution Models.

For MCA-O2S and MCA-AMS, it is a complex undertaking to identify "which advertising/marketing vehicle deserves how much credit." It requires patience and skills. And it requires your execution of multiple of the 16 strategies I've outlined for tracking online impact of offline and offline impact of online. Even more, it requires an ability (people + skills + desire) to execute controlled experiments.

So the question "who deserves how much credit" is tertiary at best.

With MCA-ADC that quest is a little bit easier. We have the multi-channel funnel reports at our disposal. Additionally in some tools we also have an ability to apply attribution models to the behavior you see in the two pictures above in the MCA-ADC section. #sweetness

The most common attribution models bundled into even the simplest web analytics tools are: Last click, first click, and even distribution.

If you are lucky, you have access to a more sophisticated tool which would include: Adjustable, based on mathematical algorithms, time decay model.

If you are among the chosen few, you'll likely have access to a digital analytics tool that allows you to create a customized attribution model.

Each of these models are applied to MCA-ADC (still without benefit of O2C or AMS) and provide you with incrementally better understanding of your digital media spend.

Each of these models comes with its own pros and cons. [If you have my book Web Analytics 2.0 please jump to page 358.] Some of them have more cons and barely any pros. Those should be avoided like the plague.

A couple of them pass the common sense test, and hence will put you in a better place than staying with last click attribution.

But most of what you'll get out of playing with these models is a deep and profound appreciation for how they'll, even in their most shining moment, give you directional guidance how to adjust your media spend (shift dollars/euros/pesos from Search to Display or from Display to Email or… other combinations).

You'll realize (even if you use the greatest customized model created by your most magnificent consultant at a equally magnificent cost to you) that success then will come not from that rough output, but rather from your ability to take that rough output, make changes, observe the impact (over weeks, or months if you are small sized), identify insights and be less wrong over time.

If you happen to be in a larger company, say you spend more than $10 million on digital marketing per year, you'll quickly see, having learned to be less wrong over time, that the question you want to answer with multi-channel attribution modeling is not "who gets how much credit" but rather "how can I optimally balance my digital marketing portfolio."

That will then drive you to seek solace in the arms of the only solution that actually works. The solution that is hard. The solution that requires unique people skills and an undying desire to scale un-imagined heights of glory. Media Mix Models. Executed via persistent controlled experiments.

When you reach that point, fame, fortune and happiness will be yours.


Multi-Channel Attribution: Closing Thoughts

This is a tough challenge. Simply because reality is complicated.

Customer experiences are ever more complex, influence channels intersect a lot more, content consumption is fragmented, the three-step "attract, acquire, retain" model is now broken into 37 different pieces.

So, you don't have a choice. You are going to have to deal with the multi-channel attribution problems, all three of them, if you want your company to have an effective advertising and marketing strategy.

Here's the good news: You don't have to try to boil the ocean in one go. In fact, that might be hazardous to your health if you attempt to do that. Take gradual steps. Increase your sophistication over time.

Here's what I recommend:

    • 1. First clarify what problem you are solving for your management team. O2S or AMS or ADC.

2. Use the appropriate set of solution (see sections above). If MCA-ADC…

3. Get really, really good at understanding your multi-channel funnel reports. They are free. They are awesome. Use the Venn diagram in the Overview report to display reality to your management team. They'll love you, and stop wasting money.

4. Start to experiment with the simple models. You are moving away from last click, you'll abandon first and even very quickly. Spend some love and attention on the time decay attribution model (ideally with several mathematical options to apply).

5. Experiment with changes in your digital portfolio based on your time decay results.

6. Measure outcomes. Go back. Analyze the data. Change some more.

7. As you master that, shift slowly to playing with media mix modeling type controlled experiments.

If at any step you notice diminishing margins of return, go back to the previous step and optimize that one some more until it is truly worth the incremental company investment to take the next step.

If you understand the frameworks, if you internalize the challenges, if you define your company's immediate unique problem clearly, and follow a step wise approach outline above you'll not just do fine. You'll be fantastic.

Good luck!

As always, it's your turn now.

Which multi-channel attribution problem are you solving in your company? Do you distinguish between the three outlined in this post? Is there a fourth one not covered in this post? Which one do you find to be the most challenging? Are you more optimistic that we'll solve AMS (across multiple screens) than I am? Which MCA-ADC attribution models do you swear by? Who's your BFF? Do you have a attribution model that's not covered in this post?

Please share your thoughts, feedback, critique, and brilliant new ideas via comments.

Thank you.


  1. 1

    Excellent post!

    As you mentioned in the post, MCA-ADC (which is realistically possible) is still mostly based on last click.

    All other methods of accurately measuring MCA-ADC need lots of data mingling. It also proves that as the organisation matures on its MCA analysis it should invest more on integration of data. (Data Marriage)

    • 2

      Sid: I agree with you that organizations will have to invest more in data marriages as they move up the "food chain."

      But before kicking off 24 month big data warehouse integration projects my counsel is to first invest time in hiring a couple of smart people with skills in design of experiments and statistics, get them to conduct controlled experiments (online and offline) and learn from that experience.

      It turns out that while controlled experiments only provide discreet results (rather than continuous, unless you are always on), they are immensely valuable results that can be applied to the business immediately.

      Lessons from the experiments can also ensure that you prioritize the best "marriages" first, :), and run away from "marriages" that will lead to a divorce at Kardashian speed.

      We essentially obsess about learning and ensure that we are making the right choices (in the right order).


  2. 3

    I love the breakdown of the three different types of multi-channel attribution!

    One additional point that seems to get missed, even with MCA-ADC(FOD), is the cognitive dissonance we often have when it comes to "multi-channel." We *believe* that, if we get sequences of touchpoints lined up, that that will enable us to assign a certain % of attribution to each channel. We implicitly strive for a "sum of the parts equals the whole" (channel/campaign A = 20%, channel/campaign B = 50%, channel/campaign C = 30%).

    At the same time, though, we recognize that the sum of the parts *should* be *greater* than the whole. Integrated messaging across channels means that channel/campaign A actually bolsters the success of channel/campaign B.

    In the breakdown above, if you removed A and C, you would expect to garner <50% of the total results. That's one additional complexifying wrinkle.

    Thanks for the great post!

    • 4

      Tim: This is a very interesting perspective. Thanks for making us all think!

      I do think that in an ideal world the sum of the parts is greater than simply summing up the individual pieces. In this case it just might be how we think about the whole.

      (Using your example) If you do A, B and C then you have an ability to reach a wider audience, and the same person multiple times (as required). That should translate into "picking up" people you might have if you did A or B or C (or just A or just B or just C), but in addition you'll also pick up people you manged to convince with messaging via A and B (and other combinations) and A and B and C. In that sense the portfolio (A+B+C) gets us incrementally more customers (online and off).

      Today we, as you say, start by saying "we got 100, who gets credit?" and that perhaps leads to a "sum of parts" thought. Using approaches like media mix models we can actually show very clearly that the sum of the parts is greater than the pieces because it delivers incrementality.

      Thinking of that does make the brain hurt. But all good things in life come from that. :)


      • 5

        The breakdown of the three different types of multi-channel attribution is awesome.

        In terms of the example above, let's treat A, B and C as sets, A+B+C = WHOLE 100%. but it's reasonable that A intersect B is not equal to empty set. In other words, some channel may not deliver incrementality but even make the whole picture stay the same. To sum up, there is overlap about their credit.

        What's matter is when it comes to "data-diriven" MCA-ADC(FOD), the overlapping in different data instance may be totally different. So how this can possibly be solved?

        Thanks for the great interesting post! I enjoyed it a lot~

  3. 6

    The quote you used at the top I find quite timely. Right now, certain people are spending too much time measuring the wrong outcomes with a mandate to justify certain legacy advertising inputs.

    Their companies would be better served to spend 95% of the time defining and analyzing the issues as they truly exist today.

    • 7

      Markers have never had more data than they have now and yet most companies waste massive amounts on ineffective advertising because they do not know exactly what drives purchases revenue etc.

      A example; you're watching TV – you see an ad and you have your laptop in your lap – you search for the brand – pull up their website – leave the website and a week later you see a retargeting ad for that brand and click on.

      Now that targeting ad brought you back to the website – you bookmark the page and then buy. The retargeting ad gets credit – the TV ad initial interest.

      • 8

        Dan: Part of the challenge is that what you describe is very hard to measure, except for very large companies or very large advertising spend. It is difficult to tie that ad view to that website visit (and subsequent view of the re-targeted ad) , and it all falls apart.

        Larger companies can use media-mix modeling (by leveraging controlled experiments), but small budgets can't.

        There will be evolution in privacy rules, government regulation, native device integration and non-currently-obvious tracking using a unique primary key across data-sets. This will help.


  4. 9
    Blair Keen says

    Another great post Avinash – the only problem is I need to take the afternoon off to read it!

    I'm curious to hear what you think the effect on MCA-AMS will be with the imminent adoption in Europe of the EU Cookie Directive?

    • 10

      Blair: On the EU Cookie Directive… It is unclear what the final set of decisions have been made by individual members of the European Union. Much of the emphasis of the guidance in the central directive seems to apply to third party cookies and less so on first party cookies (used by web analytics tools).

      That said… I'm thrilled to say that the EU Cookie Directive will have no impact on MCA-AMS. Simply because even if the directive did not exist, our problems/challenges would be exactly the same. We can't stitch (the valuable first party) cookies anyway across devices without major evolution by governments about what to do with cookie data.

      Remember none of the above applies to non-cookie based first party tracking. So if you take my example of Amazon, they should be able to do MCA-AMS with relative intelligence without cookies or analytics solutions etc based on what they collect already and store inside their systems.

  5. 11

    Makes me pine for pre-internet days. Imagine your charts w/o web, social, ipad. Simple.

    It's easy to see why sales people have lost a lot of their relevance. Their contribution is still valuable, but it competes with many, many more pieces in the overall conversion process.

  6. 12

    Thanks for the post! Very useful and it definitely makes me feel slightly better about not being able to answer all questions about all the channels and visitors/browsers and general behaviour :)

    Definitely worth a share!

    • 13

      Brad: Mission accomplished! I'm glad you feel better!

      On a serious note, you should not feel bad. This is a very very complicated area. That is why everything's so difficult. That is why simple solutions are silly. And that is why the rewards, if you put in the sweat equity, are so immense.


  7. 14
    Himanshu says

    Hi Avinash,

    Very informative post. But i am afraid both MCA-O2S & MCA-AMS models are largely theoritical with little to no way of implementing them properly even by a seasoned analyst let alone an average joe. I read your posts on tracking offline impact on online campaigns but i still find it very difficult, time consuming and expensive to implement this model with any degree of accuracy. There can be so many offline factors invloved from word of mouth, public speaking, seminars, tradeshows, print media, electronic media, outdoor media to a decision made by HIPPO that it is almost impossible to atrribute conversions to the right channels and allocate a budget accordingly. Because of these factors i highly doubt MCA-O2S model will ever be widely adopted. MCA-AMS model on the other hand is a pure theory. I think such type of attribution modelling will never be possible becuase of privacy issues. It is like putting a tracking device on your prospect once he/she is exposed to anyone of your marketing channels. I am much more interested in MCA-ADC models because this is something which can be measured at present and with relatively more accuracy than the other two models. I am really interested in knowing your views on my Proportional Multi touch Attribution Modeling: which is actually related to MCA-ADC model. You said there is no such thing as good attribution. Can you please elaborate on this.


    • 15

      Himanshu: I'm afraid I do not believe that MCA-O2S or MCA-AMS are theoretical. I work with companies everyday where we are solving these problems and helping them make incremental millions by optimizing their marketing spend across channels and devices.

      Just consider the example of Verizon included in the post. That is O2S and a real world example of what Digital Analysts/Marketers fail to consider.

      AMS is hard, but for a clothing retailer we've optimized budgets based on experiments with the use of QR codes, optimized campaign tracking etc and understanding multi device usage by consumers. It is less sophisticated than the O2S example above. But the consumer behavior is real, the impact on the business very real.

      You are right… attribution modeling is not the right answer. But multi-channel attribution is a worthy investment, just using innovative approaches.


      • 16

        Very interesting exchange of points of views! These "conversations" enrich these already juicy posts.

        I wish to add a very simple (and humble) comment: Perhaps while our society still curiously does not allow us to implant cookies right underneath the skin of our prospects, we should try and aim for precision rather than accuracy (yes, I am citing AK).

        PLEASE NOTE: This is by no means an invitation to avoid analisys. Just a thought.

  8. 17

    Love the post. Thanks for sharing your thoughts on this HOT topic.

    I have question on MCA-ADC.

    Though I have found this to be of great use to us I am not sure how this report accounts for post impression display visits?

    Also, should we be looking at it only from a month to month perspective because the cookie get refreshed every 30 days?

    I would love to hear your thoughts on this

    • 18

      Aslam: Are you referring to view-thrus? If so in Google Analytics at least the reports are showing display view-thrus. You then have a chance to decide how best to deal with that "action" in terms of attribution. If you are not using GA, please check with your vendors as some of them also show view-through data.

      The time duration depends on your business, your average visits to conversion, the mix of category (demand gen) and brand (conversion) marketing activities. My recommendation is to at least stretch the look back time period to at least a little bit longer than average days/visits to conversion. This report is a standard report in web analytics tools.

      Oh and in general first party (or even third party) cookies don't get "refreshed" every 30 days as you mention. Most will have expiry time periods of at least 9 months, many go longer. It is a different discussion if they last that long on consumer browsers.


      • 19

        hi Avinash,

        Great great article.

        I have two questions:
        1, what tool did you use to create "sunburst" visualization graph?

        2, For a product that has a first touch to purchase conversion window that is longer than 30 days, what is the next best thing to do? Does GA multi-channel report have data longer than 30 days?

        • 20

          Dylan:I use for creating the Sunburst and Chord.

          I'm afraid I'm not a big fan of first click attribution, hence I'm unable to add value to your second comment.


          • 21

            Hi Avinash,

            Thanks for sharing the link. And to be honest, I am not a fan of first click attribution neither.

            The only case we are using it is for a conversion that started with a nonbrand search and then completed via a brand search or direct traffic. We are in the personalized photo book business and it normally takes 20 days to finish a book.

            What we found is that people start their photo book project with us via a nonbrand keyword and go back to finish their project via brand keyword.

            My question: does Google's multi-channel attribution has a limited 30 days look back window? What can i do to get "click path" data beyond 30 days.

            Thank you

  9. 22

    I have to say I love MCA. I love the challenge and the rewards. I do both ADC & O2S (to some degree) and the insights are very valuable.

    I do have one issue with the Multi Channel tools in Google Analytics. I have tried many many times to work out what Google calls a first click and assist click to no avail. I have exported every conversion's pathway and hand counted all of the conversions and it never matches up. It's easy to understand when there are 3 touch points in the chain as the first is obviously the first, the last is obviously the last and the middle is the assist… but what about when the chain is only 2 steps, or 1 step? Do you have any insights into this as every model I use never matches up.

    The other issue I have is that from a search point of view, the assisted conversion report gives you the full Adwords data about a term (i.e exact, phrase, broad) whereas the conversion pathway report will only give you the search term and not the match type. This makes it very difficult for me to verify the data and this decreases my trust and the clients trust in the data.

    Anyway, just my rant about these tools in GA. Thanks for your post which shines more light on the MCA puzzle.

    • 23

      Joey: Here's the blessed gold standard page with the definitions of the three things you want: Last interaction, assist interaction and first interaction.

      I hope that helps clarify the issue, do also see the note at the bottom of the page about calculations.

      Couple other points that might help…

      First is a type of assist per GA's definition. So for a two step path, the first step would be both a first and an assist. The last step would only be last. :)

      This is small but you referred to "assist clicks," GA actually measures assisted conversions. So, for example, if you have three paid search clicks in a single conversion path – while it's true that paid search had two assist clicks, in this path it only gets credit for one assisted conversion. When trying to reconcile this with the Top Paths report, you should be sure to count conversions, and not clicks.

      Finally, great feature request on making match type available in the Top Paths report. Consider your request sent to the person at Google running MCF!


  10. 24
    Rok Hrastnik says

    Avinash, amazing as always.

    The one thing I would add, though, is that email is a great way to at least somewhat better connect the multi-device and multi-browser dots, using identifiable tracking links in email messages, to consolidate cookies/users after the click.

    This gave us some amazing insights at SM on a few levels:

    a) Helped us identify the broader impact of email on results, far far beyond just the last click (can't share the numbers, but the deferred revenue impact of email was a few times bigger than the direct last click impact)

    b) Helped us consolidate a meaningful share of users across devices/browsers and then use that to connect the dots on a larger online channels multi-channel picture

    Plus, at least in direct channels (not including retail shops and wholesale, unfortunatelly) it was relatively easy to also connect the dots between the email address (and therefore all the online source data for the user, including the actual email acquisition source) and the offline purchase.

  11. 25
    Jason Luis says


    Great post. This really is the soup du jour. Thanks for the dose of reality. I see that measurement lives on a continuum. On one end, you have media mix models which look at all possible influences yet gives you no granular level of optimization ability. One the other end, you have “attribution” models which look at a few data sources, but are less costly, timely and infinitesimally granular. The allure is obvious. =-)

    Here’s what I would do. First, decide what you are trying to do with the end goal in mind. Mine typically is to make campaigns most effective and improving ROI. Now that you got that small detail out of the way, we need to agree that the best way to do this is through optimization.

    Then you have to decide what to optimize. I think in simplistic ways. Which yields the most results with the least amount of work? Solve this, then move up the measurement maturity ladder. For direct response campaigns, you may want to apply the 40/40/20 principle. For Branding, where you’ll most likely use a survey anyways to measure success, strive to understand what messages are working best and optimize out what doesn’t.

    Side Bar – Here’s a trick that a wise colleague of mine shared with me a few years ago. Without all the fancy math of multiple regression, just put all your data into a pivot table and start moving things around. When you pull down on variable, say the site name, or creative size, note the change in response rates. This shouldn’t take long. Keep experimenting and when you see the most fluctuation when one factor is pulled in, then you have the variable which has the highest impact on response rates.

    Most things life are comprised of shades of gray and measurement is no different. When we think about multi source attribution, I think in about it as a maturity model where you ask progressively more difficult questions:

    1. Is multi-channel marketing needed? Multi attribution modeling for multiple online channels should confirm this.

    2. What source should get credit? If you’re in the conversion path, you should get some credit. If you’re not in the conversion path, well, then I would watch this closely.

    3. How should allocate spend at each level of Media Channel, Publisher/Source, Placement/keyword/creative? For an answer to this, read the post above!

    If you can get two number two, consider yourself ahead of the game.

    What not to do may be simpler than what to do:

    1. Don’t pay multiple CPA partners for the same conversion. Capture all conversions in one tool (Tag Man and Clear Saleing come to mind). You get consistency of counting methodologies and also only count a conversion once

    2. Ignore the power of planting an idea in someone’s head or helping keep the dialog going throughout the conversion cycle



    • 26

      Jason: I love the "mini blog-post!"

      I'm a big fan of a gradual evolution approach towards solving complex problems, hence you can imagine how appreciative I'm of your recommendations. Thank you.

      One small clarification… I do agree with you that Media Mix Models are not the right weapon of choice if you are trying to solve MCA between keyword A or keyword B (that's using a cannon to kill an ant). But they are indeed magnificent at strategic optimization of your media portfolio (say: tv, radio and display or search, social, email, affiliate).

      Thank you again,


      • 27
        Jason Luis says


        Thanks for your reply! I love media mix models too for the right application of desired outcomes.

        I think about it like this: The tradeoff between measurement techniques is similar to confidence intervals for statistics. "The more confident you want to be, the broader your statements [about optimization] need to be."


        • 28
          Deven Pravin Shah says

          Hi Avinash, Jason:

          Please can you shed some light as to what kind of statistical tools you use for post-processing of the data?



          • 29
            Jason Luis says


            One way to get your feet wet in fractional attribution is to use Multiple regression to build a model. There are lots of tools out there with the most basic being Excel. Free data mining tools you could try are rapid miner and WEKA. Enterprise level tools include SAS Enterprise Miner and SPSS Clementine.

            If you don't want to get your hands dirty and want a good solid tool with great visualizations, check out the newly updated Forrester research whitepaper which has VisualIQ as a leader in the pack.



  12. 30

    Thanks for another great post Avinash.

    One thing that your post-it notes with the paths drawn out reminds me of, is attribution within each channel.

    For example, taking the MCA-ADC case, say within social media, one might ask whether to attribute the social media contribution to Facebook or Twitter? Or another question might be whether the social media contribution is due to the marketer's effort in engaging with the audience or due to the amplification by some influentials?

    Understandably this is not really "multi-channel" attribution but more like "within-channel" attribution, but if the goal of measuring multi-channel attribution more accurately is to adjust media spend optimally within a unified marketing budget, then measuring within-channel attribution is basically expanding the attribution models to more granular, or segmented, views of the channels, sort of like expanding the box representing "social media" in the multi-channel funnels further to more little boxes and paths.

    This undoubtedly adds extra layers of complexity to an already complicated analysis of MCA, but if done right should end up making multi-channel attribution models easier to create and more accurate.


    • 31

      Peter: To your last point… I don't thing it makes attribution modeling easier, if anything it makes it more complex! :)

      But you are absolutely right that you want to start with strategic optimization (pan channel) and then when you master that, move into tactical, as you say, "in channel" attribution.

      At my recent keynote at the Search Engine Strategies (SES) conference I'd shown an example of how do do that. I start with a chord visualization to understand pan channel multi-channel (ADC) behavior, and then, since for that client Google was such a massive piece of the puzzle, dive into understanding behavior across keyword clusters (multiple buckets of brand, category, et al) using a sunburst visualization.

      It was such an incredibly insightful exercise!

      Thank you so much for framing the "macro and micro" in such a nice way in your comment.


      • 32
        visitor says

        Nice article Avinash :)

        IMHO, macro and micro distinction is an absolute must in attribution model of any sort. In the end, all these efforts are meant for more intelligence on customer file/segments. Naturally then, is to understand long-term implications on the business, like LTV.

        We are modeling similar to your O2S model – where each data point being on- or off-line campaign sources. For me, biggest challenge has been the matchbacks, especially for micro-level – like Google nonbrand matchback to XYZ source. It's difficult to define associations with seemingly endless permutations lolz.

        I conveniently left out the whole issue of 'justifying' the logic that drives matchbacks for the sake of conversation (or just that I am not at the level of skillset to even understand this fully :P)

  13. 33

    Excellent touchstone piece on attribution, folks would be advised to keep a link to this post handy for the next time someone wanders into their office and starts talking about "fixing the attribution problem" ;)

    If you can't measure it properly, just say so. So much damage has been done in this area by creating false confidence, especially around the value of sequential attribution models where people sit around and assign gut values to the steps. Sequence is "nice to know", it's not a measurement.

    Here's the best question I have found in reply to questions on attribution, especially from marketers: Would you be willing to kill Campaign X (or Channel X) to determine it's contribution to goal achievement?

    The answer is almost always No. Which is too bad, because one way or the other, this is likely what it is going to take to even get directional attribution, never mind absolute numbers.

    I'd argue a person who is not willing to kill a campaign to find out what it is worth doesn't really want to know what it's worth…

    • 34
      Rok Hrastnik says


      1. Totally agree on your point about assigning gut values to the steps. Was always strictly against assigning any kind of gut values — if you don't know the real or at least approximate value, don't try to assign some magical number that you just pulled out of your hat or are comfortable with.

      2. However, can't really agree with the point of being willing to kill campaigns just to figure out what's the "actual" impact. While I was still at Studio Moderna, this was often proposed (especially in trying to figure out the "real" impact of TV on online sales), but I could never quite agree.

      Information is mission critical … but the one thing that trumps it is making the monthly plan.

      I could accept the argument that "knowing" might give us better leverage in future months, and just accept an EBITDA decrease in a given period as cost … but then on the other hand it's almost impossible to "discount" all the other factors impacting your results to get a good enough picture of what happens if you kill a critical campaign.

      Let's say you decide to kill email as a channel (OK, a little extremist, but just to make a point) for a month to see what's its real impact on sales. That might tell you "something", but you can't really "discount" the past impact (from the previous month) and what impact those past activities are still having on sales today.

      Or, as was often the idea, removing the domain name from TV ads. The question is – what can you really learn (actionable insights) from doing this?

      a) The TV ads will still be driving search traffic (brand), so you'll only see a partial impact.

      b) You'll still be dealing with the impact TV generated in the past, both in brand recognition and product demand, which will still be driving people online via search and direct URL entry.

      So, measurement YES, but not at all costs, and only when there's a realistic chance you might get reliable actionable insights AND when you know you will then actually change something based on those insights.

  14. 35
    Tom Galanis says

    Great article.

    First off being in the digital space, I never realized that there were 3 definitions. Always thought it was Multi-Channel Attribution, Across Digital Channels. We are still a young industry with non-standard terms :) . Its great to see all concepts explained clearly.

    In any case, one perspective that I would like to bring into this is the customer and the if I knew question

    Attribution by channel will require one common definition accepted by all. How long do you keep the history of digital channels? (1 month, 2 weeks..). Even then having the history, does not tell you which channel was associated to the moment of truth of the buyer when he made a decision to buy.

    Before embarking down this road, it would be best to actually have an idea of the sales cycle and buyer behavior for your business (and doing multi-channel analysis in only apart of this). And also to be skeptical to the whole channel attribution concept because in the end if someone uses a channel there is a meaning to him regardless if he used other channels. Really the question to ask is, if I knew that X% used a combination of channel A and channel B before they convert, what would I do differently than before. Especially if today you are using A and B and are returning great results.

    If you are B2B the notion might take on another meaning. Perhaps as part of a buying decision on your product, multiple channels are used by different individuals in a company (based on what they prefer). One person will use whitepapers on a 3rd party site, another linkedin, another a directory, another a search engine. And they might be using this all together a a team while they are researching.

    • 36
      Rok Hrastnik says

      Tom, this might come useful:

      Especially page #8: length of purchase cycle. The "2 – 3 months before" segment is "amazing".

      So, based on this, even a 1 month tracking window isn't really long enough to get the full story.


    • 37

      Tom: There are many, complex, insights in your wonderful comment. Here are a couple of thoughts…

      I don't think there is any limit to the data you keep. You use the free Yahoo! Web Analytics or any other tool and you keep history forever. You can use a paid tool and there are many costs associated with it, but keeping data for a long time is a trivial cost.

      So the question is less "how much history to keep" and more "how long might the anonymous first party cookie (or worse third party cookie) data be of value in understanding one person." For the latter, not very long. : ) Of course you can do better if in your backend systems (not web analytics tools) you keep customer data and have clean associations for digital and non-digital touch points.

      To your skepticism… I think you are right to be skeptical about attribution models, but multi-channel marketing experience is real and something we have to understand and adapt to.

      And you are absolutely brilliant to ask: "What would I do differently than before?" That question can help lead to great initiatives and killing lame ones. :)


  15. 38

    Rok, I generally agree with your perspective. Was just trying to get to the nub of the matter, which you fleshed out – if you are not willing to give up the sales / margin by killing campaigns, then you don't really want to understand attribution badly enough!

    With the corollary: Acting on faulty models is worse than having no information at all.

    • 39
      Rok Hrastnik says

      Jim, couldn't agree more, especially about acting on faulty models.

      But since you started this topic, I'm curious. In a multi-channel setting (hours of TV air time (DRTV), telemarketing, catalog, direct mail, retail stores, wholesale, and practically all the online channels), how would you actually approach trying to understand attribution, especially considering the dozens of ongoing campaigns at the same through almost all the channels, but without discounting their compound impact and other factors (such as past impact)?

      I guess sequential channel/campaign launches could be part of the answer, but then:

      a) You're already dealing with the past compound impact of the first channels/campaigns when you add the next

      b) You need to figure in that different product categories work differently across different channels (fitness might work best on TV/online, while kitchen works best in retail, but needs TV support to generate demand)

      c) If you're just running campaigns within the same brand or launching new products within the same brand, you're already dealing with the impact of all the previous campaigns/activities, which is boosting your current results (so even if you cut off all marketing right now, you'll still be generating sales from past spill-over)

      Any advice?

      When we tried measuring the impact of TV on online sales in the past we tried a variety of approaches (calculating the online lift for brand search and direct URL above forecasted trends), plus running online only TV campaigns (which were already a problem, since they created a new setting, different from the core business) etc.

      All of these gave some partial insights, but nothing really robust enough to make future decisions on.

      Any feedback much apprechiated.



  16. 40

    Rok, the points you make above are absolutely relevant, and I don't think you are "missing" anything – there is no magic bullet. If you can't set up a control, you can't. Or perhaps you can, but the cost is so high it overwhelms the value. If it's any help, this is how I look at it:

    First, I try to understand what I can and can't measure *correctly*, e.g. can a true control be created using lists, geographic markets, etc.

    Next, of what I can measure correctly, I decide whether it's worth the cost of measuring it correctly. This is the conversation we engaged in above – cost of technology / people, cost of lost sales, etc.

    Then, if cost is too high, decide if a directional or "good enough" measurement has value, in particular, will other people understand the concept of the test, be OK with a directional measurement?

    So, for example, can I measure the lift of social? I doubt it, how do I get a control group in social? Far as I know, I can't really do a regional or market test with social, and I can't do a real media mix model, because there's too much out there already and I can't control it (turn it on and off).

    That leaves some kind of sequential media approach to launch a new product, or something similar where the control is "did not exist before". See the Nissan Leaf case study from eMetrics as a fabulous example of this approach. There, they more or less "proved" that TV can drive social activity, most people looking at the data would conclude. But exactly how much in revenue from social? Nope.

    And, the problem is, the result may be specific to the product. Really can't go any further, other than to repeat launching new products the same way to see if my results are consistent.

    But that's not "proof", it's directional, and should be evaluated as such.

    If you can't create a valid control group somehow, you really can't provide absolute numbers, and that's part of the problem with this whole topic and what Avinash is addressing – some people are representing absolute results when it's pretty clear looking at the program that absolute results would be impossible to measure.

    By the way, in my opinion, people give way too much weight to "past media" – the tail is quite short in most controlled tests I have ever seen. Sure, you can talk about past campaigns contributing to brand equity and twist yourself in knots, but ironclad numbers are not there. If you believe in that model, well, you do, and that's it.

    Nothing wrong with that, but not the same as "proof" it exists.

    • 41
      Rok Hrastnik says

      Jim, thanks for the great feedback and for explaining your process. Very valuable.

    • 42
      Rok Hrastnik says

      Jim, one more question:)

      When measuring the impact of TV on online (including past media), it seems to me that the best way for setting up control groups would be via geo targeting (unfortunatelly only in large and heavily geo segmented markets like the US, and only if you can factor in the individual geo market differences, which seams quite feasible).

      An alternative might be an online survey, but segmenting survey results based on online ad exposure (from 0 exposure to different levels of online exposure and interaction).

      What's your view on this? Or put differently, do you see any other feasible approaches for doing this?

      • 43

        The media mix modeling folks use geo controls extensively. For example, TV by itself in one market, TV + radio in another, TV + radio + billboards in another, then turn the various components on and off and see what happens to sales or traffic.

        A number of these modelers have told me when online Display is added to a market with a strong offline media presence, there appears to be no incremental gain. This makes sense to me, given what we know about the comparable media “weight” of a typical banner campaign – just not enough to make a difference if TV is roaring in the same market. This doesn’t mean “banners suck”, it means they may be ineffective when the market is “already aware” though the use of heavy offline media.

        However, plenty of studies show that offline media can drive online Search behavior, which also makes sense to me; it’s classic AIDAS funnel stuff. In other words, let each media do what it’s good at – TV for Awareness, Search for Interest, web site for Desire – and tailor creative by media type to drive desired behavior at each level.

        It should be pointed out that often one does *not* find direct channel attribution data, e.g. TV generated 50% of sales, radio = 30%, billboard = 20%. Rather, one finds what “mix” or weight for each media is most efficient for a level of sales and that is it – the effectiveness of a campaign, not a channel. Hence the name “media mix model”.

        I don’t see how surveys could replace the measurement of actual behavior, many studies and much of my own work show that people frequently behave differently than they say they will or remember. For example, the new Portable People Meter for radio is showing quite different results when compared with diaries, and the same thing happened in TV when they started directly monitoring set-top boxes instead of relying on diaries – people are simply not that good at remembering what media they have been exposed to.

        There are simply limits on what can be “proven” given various constraints, and that’s where experience and a certain amount of gut feel based on knowledge of customer kick in. If something can’t be measured properly, analysts or tool vendors should just say so, and move on to next best thing, rather than claiming absolute measurement capability.

  17. 44
    Joe Meier says


    Fantastic post.

    My company has really taken advantage of Google Analytics Multi-Channel Funnels to build our MCA-ADC attribution model. The MCF reports have recently been updated to allow segmenting by mobile/non-mobile conversions (or perhaps I only just noticed the feature). I think this could be a huge step forward for getting some MCA-AMS insight. As you rightly pointed out, tying together the same homo sapien across multiple screens remains elusive. However, I think there might be some big gains to be had by comparing the MCF conversion paths of mobile vs. non-mobile conversions. I haven't yet had time to dive into this capability but plan to in the near future.

    Segmenting for mobile was definitely on my MCF wish list. Next would be a further split by tablet/phones, and the ability to see time when looking at conversion paths. In other words, how many days passed between Visits 1, 2, 3 and 4? That would go a long way to help build some of the time decay models you mentioned.

    This post is definitely getting bookmarked. Great insights.

    -Joe ('Joe from Illinois' from our recent Web Analytics TV interaction)

  18. 45

    First of all, thank you for giving such a deep explanation on MCA!

    Now I'm excited to learn more about Analytics (currently going through the IQ modules). Hopefully, I can benefit my organization by showing the real profiting channels with MCA.

  19. 46

    Avinash, great post & timing.

    Curious about how remarketing flows into this mix. We see direct conversions from Remarketing and of course view throughs (assisted) but beyond that?

    How do you measure what percentage of direct, organic (online) or visit to store and/or phone calls (offline) are driven by re-marketing efforts?

    Thanks again!

    • 47

      Bibi: Remarketing obviously comes into place in the context of MCA-ADC. It is one of the key strategies to do advertising that is much more closely tied to the intent of the user (they already know you, you already know them).

      In context of attributions… you capture a remarketing touch point just as you would any other touch point, be it Search, Display, Email or Social. We just have to make sure that the ad is tagged with the proper metadata to allow for it to show up properly in the Multi-Channel Funnel reports (and view thrus and clicks show up there for display).

      Once the data is captured, we have an ability to do analysis and rest of the stuff outlined in the post.


      • 48

        I think remarketing is really tricky to measure, especially if you're paying CPA commission to the advertising agency.

        I mean, if I should pay for CPC and Display conversions XX% per conversion, I would pay a lot less for a remarketing conversion when there is no other advertising campaign (managed for by the agency) in the MCA-ADC within the 30 days.

        Is this reasonable or I’m crazy?

        • 49

          Juan: Why would you not pay commission, or some type of fee, to your agency for any other kind of work? For example, you pay for CPC (to your Agency and not just Google), you pay for social campaigns, etc. etc. So in that sense your CPA commission for remarketing is not such an odd occurrence.

          In context of attribution, to get to your reasonable or crazy part, remarketing is just another touch point for the customer. It will be likely closer to the conversion point, it will likely mean you understand more about the customer, but you will see it in your MCF reports and be able to handle it just the same as other channels when applying attribution models.


  20. 50

    You're always giving me more to think about, Avinash. Thanks for that. Should all companies be thinking about attribution, or do you see it as a concept that companies should tackle only after they've demonstrated proficiency with other analytics concepts?

    I see a lot of folks voicing interest in MCA, but they haven't even properly identified their KPIs, much less valid metrics. It seems like MCA isn't something people should even be concerned with until the basics have been covered. Is that the right way to think about it, or should people pursue the critical basics and the advanced stuff in parallel?

    • 51

      Josh: I'll go out on a limb and say that there are a long list of immediately impactful easily achievable things that companies should do before wading into multi-channel attribution analysis. Evolution, not revolution. Mostly because the former leads to greater sure success over time.

      There was some guidance in my post on who should consider attacking the problem first (after above). Companies that spend more than $10 million on online advertising per year. That is not a hard and fast number, more the kind of scale at which you have to be to have this be a problem worth solving, to commit to solving this problem, to have the skills in the company to be able to find a solution.

      Fragmented customer experience is a reality. It will become a larger part of the influence ecosystem over time. Everyone has to solve for this higher order bit.


  21. 52

    Also don't forget the realpolitik of budget setting FMCG style where risk averse rules:

    1. Many Marketing budgets are set 6-9 months out and NOT planning to spend it is either a sign of incompetence or risks it being grabbed by finance.

    2. KPI: Marketing Directors get fired for losing share (but gaining TOO much share is a big no-no as this resets targets for NEXT year .. which if you miss, yes you guessed it, you get fired). So Goldilocks rules…..

    3. LT brand share is a function of brand values – therefore only a stupid Marketing Director will voluntarily axe his budget based on AMS channel analytics unless the alternate channels deliver proven long term brand effects even if short term sales could be inferred to channel specific pressure ..

    4. But they will pour money into experimentation and thats where a lot of the pressure on MCA is evolving from and why I think clients re requesting the holy grail even though tongue in cheek they probabaly know its impossible as Avinash says.

    5. Therefore High levels of digital experimentation (makes us look good ! ) and flight to safety under budget pressure.

    6. Incremental knowledge is good, constant learning and improvement, the system is not good at adapting to big allocation changes which impact budgets tooooo much at once as this raises the risk profile for the budget holder.

  22. 53

    So well though out. What a great post.

    My personal frustration with MCA is it can often ignore non-marketing factors. Why, for example, did the O2S controlled test fail in certain markets? What? The people in the stores were clueless about the products and had terrible breath? Isn't it possible those factors made more of a difference than the media in converting customers!

    Businesses look at marketing budget distribution as "easy," while they look at the operational aspects as "hard." And this does make sense. But when businesses spend millions of dollars on tests, statisticians, economists, databases and custom software that maps weather patterns to buying habits, sometimes it's actually a lot easier to go back to the basics of your offering itself, not just the "marketing."

    Marketing, as a word, seems bastardized. Verizon, while it does have a marketing department, is a marketing company as a whole. Advertising and marketing seem to be synonymous today, which seems to lead to businesses siloing the process of making promises (advertising) from the process of keeping them (operations, product, service, etc.)

    I guess what I'm trying to say is that it has always seemed easier to optimize a "marketing" budget across and within channels, but is it a possibility that may businesses, in fact, are in a place where it's more cost-effective (and more potent) to look at the offering itself?

    And this isn't to say it's an either-or proposition. Both need constant attention. But I feel we, in our practice, advertisers in theirs, and businesses in general point their finger and their efforts disproportionately toward the fuel going into the engine, rather than examining the engine itself and its ability to turn perfectly good fuel (prospects) into motion (customers and cash flow).


  23. 54

    Avinash thanks for this eyeopener post, I knew about MCA but you have added fresh perspective with O2S,ADC, AMC, ADCFOD.

    Just wondering if you have some thing MCAADCFODFOB which is like Multi-channel attribution across digital channels for one device for one browser.

    • 55

      Arpit: In case you were serious in your comment… What you are describing is MCA-ADC – the third one described in the blog post.

      Solutions like Multi-Channel Funnel reports in GA (and similar reports in other tools) measure exactly what you are describing, user behavior for one device and browser across digital channels.


      • 56

        I am little unclear on browser part let me re-phrase my concern so let's say in my laptop I have chrome, firefox & IE9 installed and I randomly use any of these 3 browser types now my understanding is that cookies are browser specific so just like you clarified MCA for one device on same lines I am guessing multiple browser usage on same device also creates a limitation for any web analytic tool, correct me if I am wrong like do these tools have ability to create cross browser readable cookies??

  24. 57
    Anthony Power says

    Thanx for all the commentary and discussion – had to read everything at least twice.

    One of the complexities we face is not only the channel, device distinction but also the type of content served. For instance communal content like recommendations can be both on a social channel as well as reviews on the web site; and emotional/aspirational content can be television or organic search as can promotional content.

    Since content is likely to influence decisions shouldn't we also be thinking about a structure to categorize/tag content?


  25. 58

    Avinash you've completely blown my mind! Thank God you put those post-it diagrams in otherwise I'd have been onto a loser! Well that has to have been the most comprehensive post I've read in a few weeks!

    But I have to say a big thank you for doing your best not to be too techie with explanations. I don't think I'll ever get to grips with the measurements but now I have a much clearer explanation of the 3 variants of MCA – before today I'd never heard of them! It's true your posts are serious fountains of knowledge!

    Thank You

  26. 59
    Todd Shelton says

    Hi Avinash,

    It looks like you may have flipped the Verizon results–they report 5 in-store activations for every 1 online activation. They appear to have tested increased generic Google keyword spend against in-store activations, though that's not super-clear.

    Great article and thinking. Love your sticky note illustrations–can I please steal that idea? :)


  27. 60
    Harpinder Sohal says

    Hi Avinash – Most of the comments here have echoed my appreciation of your post on MCA, but at the same time has made me realise the complexity of it! The best part of all for me has got to be:

    "Next time you hear someone utter the words multi-channel attribution, the single greatest gift you can give yourself is to ask in your sweetest possible voice: "Are you referring to MCA-O2S, MCA-AMS or MCA-ADC?"

    Now I've got my head around MCA-O2S, MCA-AMS & MCA-ADC I can confidently ask the question next time the topic of MCA is brought up.

    I've looked at multi-channel funnels in GA, but was wondering if you what other online tools you would recommend that offer similar or more detailed level of insight on MCA?

    Appreciate they won't tell the full story but was just wondering if anything could start scratching the surface.

    • 61

      Harpinder: I'm glad you found the post to be of value.

      As I'm sure you'll notice when you log into Google Analytics much of MCA is about clean data collection and campaign tracking parameters and other such things. So it is less a tools problem.

      But there are many tools that allow you to see this data beyond the single session view that most web analytics tools limit you to. In addition to GA you can see pan session behavior in Adobe Discover, IBM's CoreMetrics etc. You can also explore specialized tools like Tagman and ClearSaleing and others.

      Whenever you talk to a tool vendor (all of the above and more) you should ask them: "Come on, what is it that you really track? O2S, AMS or ADC? And what do I have to do to make it work?" Fun discussions will follow. :)


      • 62
        Harpinder Sohal says

        Thanks for the reply Avinash.

        It's good to know that there are tool vendors thinking more and more about MCA.

        I came across this article about Google's move to "Active View" for ad impressions – – certainly adds yet another perspective!


  28. 63
    Dave Rekuc says

    Great article. I must say, I was drawn in by the allure of MCA-ADC, but eventually realized complex models that don't drive insight or value are just complex for the sake of complexity.

    Regarding MCA-AMS, its funny that I just saw this post today. Google is partnering with NBC to do a study on how people are interacting with the Olympic games across multiple devices. Hopefully this study helps shed some light on we engage multiple channels.


  29. 64

    Hmm…MCA-AMS, that's a headache!

    Immagine you had the same browser across multiple devices (laptop, pc, tablet, phone) that could sync your traffic across screens when you are logged in. Immagine it could pass cookie information too.

    Would you sample the traffic behavior of said browser (let's call it browser X, no better, browser C!) as a proxy for the behavior of your customer base (including users without the same browsers across devices), or would you assume that the ability that the browser gives the user to share content across his devices would also increase the likelihood of cross-device navigation (and therefore wouldn't represent correctly the population not using browser C)?

    Thanks Avinash for the great post!

    • 65

      Paolo: To the best of my knowledge what you are describing is not possible in the context of data capture into web analytics tools.

      But depending on the sample size, and sample type, often it is ok to extrapolate behavior. You can also complement that data with other segmented analysis to validate your hypotheses, or use experimentation for the validation.


      • 66

        Avinash, this is the first time I have read one of your posts, great content and comprehensive.

        I find MCA-AMS interesting. iPad and iPhone now have iCloud tabs – is this going to help enable tracking over multiple screens?

  30. 67

    Part of the reason it makes sense to spend time defining the problem, is perhaps simply because that's all about the information, and it takes time to take that in.

    While the kernel of a creative solution is something that the brain almost trips over — can happen in an instant.

    So the time spent defining the problem is food for the right solution to happen, as well as gestation time.



  31. 68

    I am just curious. I read over your article some.

    I am wondering… isn't the main comprehensive solution going to control of the browsers?
    Google is quieting suggesting, "you don't know what you're missing" browser login.

    Once that is mandated or highly incentive-ed, doesn't that provide the possibility for a more comprehensive view into mult-channel attribution?

    Just my thought.

    • 69

      Tom: I do not believe this solves the problem across browsers. It most definitely does not solve the problem across browsers across devices.

      If you are logged in, as I covered in the post in the example of amazon, then you browsers or device level controls don't matter – the site is collecting the data regardless and can use to gain an understanding of multi-channel behavior.


  32. 70

    hi kaushik

    you say that mca-adc does not allow for mca-o2s but isnt it true that lets say if you had a loyalty program of some sort (e.g. best buy's reward zone) then you can tie together the activity online to the same individual who made the offline purchase in the store and you could then do mca and o2s in one go?


    • 71

      Niren: There are two scenarios when it comes to measuring the effectiveness of all your digital channels, individually identified, in terms of driving offline sales.

      The first option has three steps….

      1. Meta data from the controlled experiments you are running across all digital challenges. 2. The ability to tie that data to your offline data. 3. Analyze all that offline to measure the impact on test versions vs. controlled versions.

      Or a second options is…

      If you can identify by PII exposure of your digital advertising (you can know my name when I see your ad on Google or Yahoo! even if I did not visit your site) and then know when I walked into the store and bought something and tie those two things together, then you might also be able to measure the effectiveness of digital advertising in driving people to the store.

      Typically the option immediately above is not a possibility due to privacy and internet data collection structures.

      I'm not sure where you were considering the loyalty card data, but I hope the above clarification helps.

      PS: You'll notice I'm ignoring the obvious: If you have a Best Buy loyalty card that you are then incented to log into every time you visit, no matter which digital channel you visited it from, and then you make purchases in the store that data can be tied together and analyzed. This is the amazon scenario I'd shared in the post.

  33. 72

    This is a great piece of work that really helps to clarify the multi-channel challenge.

    Thanks Avinash, ground-breaking clarity.

  34. 73

    Hi Avinash,

    Many thanks for this excellent post, it's helped me to understand the three different types of MCA – you have a great way of explaining things.

    I've been developing my own MCA model for the ADC type of attribution. I'm using the data I can get for free from GA's Top Paths report. The model I've built in Excel segregates the path data into seperate sections and assigns each path step some weight of the conversion value. The weighting model is flexible, but not advanced enough to get to the customisable algorithmic functionality as of yet (or the time decay version). It curently handles first-click, last-click, first&last-click and equal spread.

    The results I see when I plug the data into the model are interesting. Normally, I am seeing that the Direct channel gets more weight once filtered through the model than it gets in GA and subsequently the marketing channels get less… I understand that GA will assign the conversion value to the penultimate channel touch-point if the final touchpoint is Direct, wheras my model will not discard the last click as direct.

    e.g. SocialNetwork>OrganicSearch>Referral>PaidSearch>Direct will assign the value of the conversion to PaidSearch, not Direct.

    I suppose this should be expected as the Direct channel is essentially a converting channel, whereas paid search and organic are assisters (or introducers).

    So with my MCA-ADC showing the marketing channels (Organic, Paid, Email etc) in a worse light (or less revenue generating light) than GA shows, I wondered what insight I could take from this to actually turn the information into a constructive plan or strategy. Have you seen this happening yourself?

    • 74

      Rob: I congratulate you on your efforts, it is so amazing to read about your efforts to make sense of this obviously complex issue.

      Let me separate two things…

      1. When you look at the standard reports in GA the conversion credit does go to a campaign (paid, seo, email, social, etc) if the last visit was via Direct but there was a prior campaign visit.

      2. The reports in MCF show the complete page, including if Direct was the last one. The newly released attribution modeling tool in Google Analytics will allow you apply attribution models you prefer to that path. At the moment you can do First, Last, Even, Time Decay (with built in flexible algorithms), and custom models. It is nice to be able to do all this inside Google Analytics itself and look at the shifts in attributed value in terms of credit for each channel.

      The challenge is trying to figure out a structure to attribute the aforementioned credits. There is no standard, and the distribution, as you can well imagine, is open to each person's biases. :)

      My recommendation is to look at the shifts (differences that occur when you move from Last to First or First to Time Decay etc), create hypotheses on the optimal media portfolio, test and iterate based on the results. I realize this sounds like a lot of work, sadly the problem is quite complex.


      • 75

        Hi Avinash,

        Thank you for your reply.

        I did see the update from GA about the new Attribution Modelling tool, but it seems it's Premium only (as far as I could tell). At this time I'm not one of the "chosen few", as you put it, who have access to an advanced attribution model, so I had to make my own!

        Your suggestions are exactly what I've been trying to do, which is reassuring! I've built my Attribution Model to be able to compare two different model weightings together, as you detailed above. Looking at the results shows me how, in some cases, Paid Search will be 'more valuable' in a First Click scenario than a 'Last Click' scenario which says to me that the PPC is bringing in lots of new customers and is the entry point of many conversions. (Or at least is providing more touch points at the beginning and middle of a conversion path)

        The challenge is to now hypothesise as to what these differences in results mean holistically to a business and then test media portfolio options.

        I love a complex problem and this one seems to be particularly thought-provoking.

        • 76
          Joe Meier says


          I have also headed down the path of a customized Excel-based attribution model for my company. Our approach sounds similar. I'm using calculations off of the MCF Top Paths report after having created a customized Channel Grouping in GA.

          I found your discussion of Direct particularly interesting since that is right now the most impactful assumption I play with. I've tweaked my model for first-click, last-click, percentage distribution, etc to observe the effects, but the most important assumption involves how I deal with Direct and Branded Searches. Assuming proper link tagging (and thus Direct data that is as clean as possible though still imperfect), I think Direct and Branded Searches can be treated very similarly, namely, as people looking for a particular website (navigational intent). Direct certainly involves a component of "don't know" too.

          Because my business is purely online and not a nationally well-known brand, I ignore Direct/Brand when assigning credit for conversions. In other words, I would treat [PPC Ad] > [Brand Search (paid or free)] as credit goes entirely to the PPC Ad. This leaves some transactions out that only ever have Direct or Brand in their path. One could argue that those should be re-distributed among the other sources, again, making the huge assumption of a pure online business with all links properly tagged.

          I recommend you look back in your Google Analytics MCF data to around 10/28/2011. I had an extended email conversation with a GA support person a few months ago because I observed a huge spike in Direct interactions occurring at the end of conversion paths in MCF (not in other GA reports, which makes sense with Avinash's explanation). This turned out to be due to a technical correction made in GA at that time that started recognizing more Direct visits that were previously missing. It had no impact on my model since I "ignored" Direct anyway, but it sounds like it might have a huge impact on your model.

          Curious what your thoughts are. I also love a complex thought-provoking problem – with huge marketing budget implications of course!


          • 77
            Dave Rekuc says

            Rob & Joe,

            Some very interesting points on MCA:ADC. I agree with Joe regarding ignoring the last-click on brand terms to allow the penultimate touch to get the credit, just like with Direct. Not sure if it was mentioned, but custom channel groupings in the MCF section are really useful for this purpose. I like to group display, non-brand ppc, branded organic, etc using the custom channel groupings.

            My own personal warning regarding MCA:ADC, is to make sure the analysis you're doing has enough potential impact to justify it. I would take a look at channels/campaigns you think are undervalued and see how much room there is for it to be "turned up." I've gone through a lot of this work before to discover some undervalued campaigns were already basically maxed out (ad position 1, etc).

            Personally, I was drawn to MCA:ADC because its a fun mental exercise, but truthfully, I've found more value in investing my time in customer level metrics (12 month value, CLV, etc) and applying it to our marketing budgets/goals at a granular level.

            Just a friendly warning, because MCA can be a huge time sink and if it doesn't yield fruit then all it did was divert your efforts. Obviously, every business is different and it may be a more pressing issue in your organization. Unfortunately, we have finite resources and need to allocate them where they'll have the biggest impact.

            That being said, this topic will always fascinate me :)

          • 78


            It's good to hear someone else has taken the plunge into home-made attribution. I had thought about ignoring Direct / Branded search in my models, but wondered if it would be better to keep it in for a more transparent picture of how Direct impacts on the overall figures.

            As a Search Marketer, I would ideally like to see more value attributed to channels which we have a direct impact on, rather than it being attributed to brand – so I see the allure of choosing to ignore direct last click.

            As Avinash says, each model is subject to personal biases based on our interpretation of 'what's going on here?' so I think there will never be one answer that everyone can use, only everyone's best guess at being closer to the truth.

            Thanks for noting the changes to GA around October last year, I will check and see if there were any significant differences before and after.

            Dave, thanks for the warning and I can appreciate that it's a big time consuming task. But, like you, I have an inquisitive interest in MCA which may prove to be extremely beneficial.

  35. 79
    Jonathan says

    Hi Avinash, I have only just caught up with this months posts. I have a multi-channel distribution problem at a very base level I would ask your opinion of especially with the latest Google updates in mind.

    I have a small number of short TLD's in Polish. The translations would be CarInsurance MotorcycleInsurance LifeInsurance HouseInsurance under the dot com code / brand. I am in the process of developing a aggregator / price comparison platform.

    It is my intention to develop four individual, websites under the keyword root domains each being authoritative to its subject URL. This offers me the possibility for multichannel distribution.

    Is it permissible, in your opinion, to use the same price comparison platform for each website ? or should I single out the comparison platform for the individual insurance cover specific to e.g. Life Insurance in isolation from the entire comparison platform option?

    I ask the question as I do not want any duplicate content problems even though the platform is a service to the public.

    • 80

      Jonathan: I'm afraid this is a bit beyond my core area of expertise and hence I don't have anything of value to share.

      Perhaps hiring a good SEO consultant might be optimal.


      • 81
        Jonathan says

        Thank you 4the reply Avinash, The truth is the very top SEO's are out of my fee bracket & soooooo many others are just wear the emperors new clothes.

        • 82
          Arthur says

          Hi Jonathan

          I have worked on a number of aggregation websites before and common problems that I see are silos where different sites are run on separate platforms and gaining insight into customers across multiple sites becomes difficult.

          Ideally what you want is to understand the commonality between the various sites and implement these on a common platform. Analytics and tracking should be common to aid understanding of the customer.

          Next identify where the business needs to be able to flex (its critical that this is led by business flexibility rather than technical flexibility – its often done the wrong way round). Make sure that the platform can flex based on the business differences between the sites (do the sites need different question sets? are the calls to action different on different sites? are some sites CPC and others CPA? how is the product information for each site gathered?)

          This should give you the flexibility to make each site tailored to its distinct niche while at the same time having one platform that serves (and persists) the common elements.

          Hope this helps

          • 83
            Jonathan says

            Arthur@ Thank you, really appreciate your input.

            You are absolutely right re flexibility to make each site tailored to its distinct niche. The best scenerio would be " having one platform that serves (and persists) the common elements." This would allow for the multi-channel access and importantly, to co-brand the keywords. Quinstreet, Bankrate, Lendingtree verticals are the models.

            My generic mail is wanda @ live . co . uk If you have knowledge of the European markets do get in touch.

  36. 84

    I have been using the Multi-Channel Funnels in Google Analytics to help update our Revenue Attribution model. While working on this model, I have been comparing Multi-Channel Funnel revenue to revenue found in the Traffic Sources report. Overall Revenue for our store always equals out between the two reports. However, when I look at revenue by medium, my direct revenue is always overstated in the Multi-Channel funnels report.

    My assumption is that revenue for all paths that contain only (none) in the Multi-Channel Funnel report should equal the total (none) medium revenue in the Traffic Sources report. It does not. In the Multi-Channel funnels report, I am only adding the paths that contain (none) as the medium, meaning paths like (none), (none)>(none), (none)>(none)>(none) and so on. I am comparing this to the total direct traffic listed in the Traffic Sources report. I understand that the Traffic Sources report disregards any direct touch if there is another medium touch before it.

    I have been able to segment these down to find the individual transactions that are causing the difference. It looks like certain transactions that are showing up as Email, Referral, Organic,Paid in the Traffic Sources report are showing up as Direct in the Multi-Channel Funnel report. This is causing an overstated amount of Direct traffic in the Multi-Channel report.

    Has anyone else run into this same problem or does anyone know why Analytics would be tagging the same transaction as different mediums within two different reports?

    Any help you can provide on this question would be much appreciated.

    • 85

      Tim: From a macro perspective in all other reports if there was a campaign present in the customer's journey prior to a direct visit then that campaign is credited for that conversion – even if direct is the last visit. A campaign in Google Analytics is SEO, Email, Referral, AdWords, AOL Banner Ads etc etc.

      In the MCF report the logic is a bit different, direct gets looked at with a different logic and gets credit for the last. Often that causes the differences.

      Which one is more accurate?

      I personally prefer the MCF treatment better. I do want, if I'm using the last click attribution model, direct to get credit if it was last.


  37. 86

    Very informative post Avinash!

    I've one question "How to Track Affiliate Revenue in Google Analytics"?


  38. 89
    doug little says

    Attribution is a bunch of crap invented by marketing folks to justify their enormous spend. People don't buy products because of who or how they were referred to the product – at least not people with a brain.
    If product development and support got 1/2 the funds of marketing, a company would have a superior product and people would buy it.

    Why don't we spend our energy on making better, cheaper, more reliable products.


    • 90

      Doug: Ha, ha! Nice comment.

      I wish it were that simple.

      A lot of stuff about attribution out there is truly high quality smelly stuff – to sell consulting services or expensive software. And we all have to train our radars to be able to detect that.

      But it is equally true that with media fragmentation, evolution of consumption, the reality of our always connected world with multiple devices, cause us to live in a very complex marketing scenario. So we have to understand the consumer journey, and more specifically what journeys result in positive outcomes for our company. One of the surest ways to understand this is to leverage controlled experiments (rather than then let's distribute arbitrary credit).

      The end goal is not to figure out how to spend more money, rather the end goal is to figure out how to spend the existing budget optimally. Which ugly flowers to stop watering, which pretty flowers to water more. :)


  39. 91

    Dear all

    I was reading through this post (and actually thumbs up for it) when I had a thought, so I just want to share my 50 cents with the community:

    When talking about campaign measurement, MCA-XXX, campaign value/ROI we always think about positive contribution/assists.


    Brand campaign A is seen by 2000 unique browsers, 500 of which finally respond to email campaign B and convert.

    We can infer that campaign A assisted 0-500 times conversion (precise figure to be found by analysts) But what about the other 1500? Did they just not convert because not interested, or were they discouraged by campaign A to do so?

    Maybe conversion funnels should have dead-end-leafs, designating the number of visitors stopping interaction with the brand after last click…

    Best regards

  40. 92

    Hi Avinash,

    I'm a big fan of your blogs. They're so in-depth and helping me so much in my job.

    Anyways, is there any reason why you only list 2 blogs per page? I find it hard to look for some of your previous posts sometimes when I need to look up some tips.

    Again, thanks for your wonderful tips and advice.

  41. 94

    In the begging you say that you need to spend 95% of the time trying to find the problem, but then at the end you challenge us to find the optimal marketing portfolio.

    I don't think that you can find the optimal portfolio without having an agreed attribution strategy. I recognize testing, but I think that testing can be avoided by using a "causation/influence" model. BUT, you gotta have the data to do it. Fortunately, my company does.

  42. 95

    Hi Avinash

    Tks again and again, i read your article, some of the comments . (still hard for a french guy to keep my brain up reading english) ..

    What's your opinion regarding the google universal protocol (beta), Is it the new graal for the model MCA – AMS ? Does it infringe the policy cookies ? As i don't use it for the time being, ..Are the principles of the protocol based on multi-session cookies ?

    Tks a lot !

  43. 99

    Good read.

  44. 100

    Hi Avinash

    This is my first read of your blog and found it to be thought provoking and informative. Thanks for sharing your insights. Wanted to share my 2c.

    As a Statistician, I think that postulating hypotheses of purchasing behavior and conducting experiments to test the hypotheses seems to be the most prudent approach as we can build on insights leading to better understanding of purchasing behavior.

    Also before building models, spending time to explore the different facets of the digital path leading to conversion is also vital.

  45. 101

    Dear Avinash,

    Thank you for your great post! Just one question please.

    I’m wondering, why through Multi-channel funnels I can see let’s say 14 conversions while from All Traffic – Ecommerce tab active – filtered by Source: {one of my referral websites} I can see only 8 transactions?

    Which data I should believe? Why there’s such a difference?

    Thank you!


    • 102

      Alex: This is a complicated question, but let's try and super simplify.

      All other reports in Google Analytics, except MCF, use an attribution model called Last Click Non Direct. Under this model a direct visit does not get credit for a conversion if there was any other source for a prior visit by the same person (organic, email, social, ppc, affiliate, anything).

      MCF uses an attribution model called Last Interaction. Whichever source was last (including direct!) gets credit.

      So when you look at the conversions this will cause your numbers to be different. Yes, in an optimal world the Google Analytics team will not torture you with this type of implementation. But such is life.

      If you want to reconcile this data, in your case referrals conversions, it is pretty easy. Go to the Multi-Channel Funnels section in GA. Click on the Attribution Modeling Tool. This is free for everyone. You will see that the report shows All Traffic Sources, and applies the Last Interaction (also known as last click) attribution model. Right next to it you'll see a link called Select Model. Click on it. Choose Last Non-Direct Click.

      Now your table will have two columns. When you look at the column titled Last Non-Direct Click, you'll see your referrals show conversions exactly as all other GA reports. When you look at Last Interaction column you'll see your MCF data.

      Boom! You are happier. :)

      PS: In case you are wondering which one is right…. I prefer the MCF attribution model, Last Interaction. I believe that all other GA reports are wrong to use Last Non-Direct Click attribution modeling – they need to be fixed.

      PPS: In the small chance that you don't have access to Attribution Modeling Tool in your GA account, please fill out this form

  46. 103

    You just gave me a new trick to become more productive.

    Thanks so much.

  47. 104

    This is really helpful.

    Thanks for sharing!

  48. 105
    Gwen W says


    Great work on this post. It is very informative and detailed.

  49. 106
    Avinash says

    Fantastic post!

    Just what I was looking for to get a grip on the problem. The differentiation of MCA in to three categories really helps me to probe my problem better. If you (or any other reader) gets a chance to reply; is MCA discussed in your book?

    Also any further help/pointers is appreciated.

    Avinash K.
    (Umm,,, yes my name and initials are the same as the author, but I am a newbie:))

  50. 107

    I can't get over how this article that was written so long ago is STILL so relevant. What I'm seeing more and more with all of these really expensive analytics platforms is that none of it is hard science – it's all theory. They are taking their best guesses when it comes to AMS and O2S and sometimes even ADC!

    The only people right now that can really go AMS are platforms that require login to see the site (i.e. facebook and to a degree LInkedIn and Google). The more friction they create the less users will want to use them.

    The ADC attribution is really the only thing you can think about doing. I like the idea of custom configuration and time decay. I still think that for small businesses you can use first, last or equal attribution models since their paths are so short. I'm actually surprised there aren't more companies focusing on simple lead attribution for these segments. I assume it's because these are not clients that will spend thousands of dollars a month.

    The whole conversation is fascinating. I'm glad I found this blog.

  51. 108

    Hello Avinash and readers! :-)

    Granted, this is an old post and it's littered with comments of "dubious quality" at the end but I thought I'd add still mine.

    While working with Croatian National tourist Board I have encountered all of these issues and more – there is no direct conversion here since all we do is "inspiration" and "branding". No sales, some downloads and outbound clicks but mostly, that's about it. Our mission however is to "bring the traveller to Croatia". Period.

    The challenge here is measurement of effects optimisation of a highly complex media mix. Offline with multiple publishers, TV, Radio, outdoor and online, all major PPC channels, social, display advertising…. Multiply that by 20 worldwide markets, multiple platforms, devices …you get the picture.

    What I am thinking of is sort of a framework. How many tourists visit during the year if we do (or don't do) this? What happens if we completely turn of our Adwords advertising? How about display campaigns?

    Tool which I have been obsessing about for a while now are surveys. Simply ask all the tourists in Croatia have they visit our website. How much did it nudge them towards purchasing the Croatian story? Have they seen our campaigns and what they think of them? 4Q and SurveyMonkey spring to mind, as well as Google Surveys.

    Any tips feel free to share. This is a gargantous effort and a serious challenge.


  52. 109

    This magnificent article will soon be 6 years old and the advanced attribution has not advanced as far as we thought it could have.

    In Spain, there are few large companies that have given way to this type of attribution and Google is delaying its rollout.

    I hope it will soon be something more widespread and used by a greater number of companies.

    • 110

      Antonio: Here's something I did not realize enough, and hence did not stress, enough when I wrote this post… Attribution is not a technology problem, it is a politics problem.

      The reason we have not made more progress is that we have not addressed the political dimensions. What are individual teams incentivised to do? What structural enticements exist to solve for the global maxima? How does the company balance for short-term and long-term outcomes? In most companies, teams solve for their silo only over a short period of time because of how team bonus is computed. This is a hard organizational problem to solve, hence most companies fail at attribution.

      I am optimistic though. New machine-learning algorithms solve for the attribution outcomes at a scale that was unimaginable AND action can now be automated which to a degree solves the politics problems.



  1. […]
    Multi-Channel Attribution: Definitions, Models and a Reality Check,

  2. […]
    Avinash Kaushik posts "Multi-Channel Attribution: Definitions, Models and a Reality Check" at Occam’s Razor.

  3. […]
    Multi-Channel Attribution: Definitions, Models and a Reality Check,

  4. […]
    Avinash Kaushik has some data goodies to start off our week on his Occam's Razor Blog: Multi-Channel Attribution: Definitions, Models and a Reality Check

  5. […]
    Avinash Kaushik published an important blog post last week entitled "Multi-Channel Attribution: Definitions, Models and a Reality Check." It's a valuable framework for understanding attribution, which is a fancy way of saying "how we understand the impact of marketing and advertising content on sales." If you're in marketing–or if your business depends on digital marketing–you need to read it.

  6. […]
    Multi-Channel Attribution: Definitions, Models and a Reality Check. Occam's Razor

  7. […] Multi-Channel Ascription: Connotations, Designs plus a Fact Verify ( […]

  8. […]
    Muito provavelmente, não. A atribuição de valores por canal é um assunto muito delicado e cheio de detalhes mais complexos do que a distribuição por média aritmética permite cobrir. Se você estiver interessado em desbravar esta área, eu sugiro começar pelo artigo (em inglês) Multi-channel Attribution: Definitions, Models and a Reality Check, do Avinash Kaushik – no artigo, ele esclarece bastante sobre o assunto, especialmente sobre a complexidade envolvida em se fazer a atribuição/distribuição de méritos entre os canais de conversão.

  9. […]
    A veces, el concepto “modelo de atribución” crea confusión, ya que no se usa sólamente para hacer referencia a los canales online. En ese sentido, Avinash redactó un artículo sin desperdicio que lleva por título Multi-Channel Attribution: Definitions, Models and Reality Check en el que diferencia tres modelos distintos de atribución multicanal y las posibles formas de medirlos: de online a tienda, multi pantalla y a través de canales digitales.

  10. […]
    Avinash Kaushik, Multi-Channel Attribution: Definitions, Models and a Reality Check, Occam’s Razor, 2 Apr 2012. Discusses how to figure out how much business is attributable to various channels of advertising.

  11. […]
    Most beginners think social media measurement is about counting the number of eyeballs. But what really matters is the actions those people take as a result of your social media efforts. This can also be known as a conversion. A conversion means taking action, whether it’s clicking a link, filling out a form or making purchases on your website. It all depends on your organization’s goals. For more information on conversions, check out this post by Avinash Kaushik.

  12. […]
    Some of the industry’s thought-leaders are already starting to distil these new user tendencies and are suggesting that the holy-grail of modern, mobile analytics is to accrue data for multiple interactions across multiple screens. Avinash Kaushik refers to this as “Multi-channel Attribution, Across Multiple Screens ”. This is the idea that we will be able to track multiple interactions from TV to Tablet, Desktop to mobile. MCA-AMS would allow for the analysis of a full research-to-buy cycle, tracking customers switching between devices by using services such as Evernote and Pocket to keep track of user interests. The future of an MCA-AMS system is still in the “blue-sky thinking” stage, but with many of the industries greatest minds thinking about the issues; surely it won’t be long until there is a solution in place.

  13. […]
    Avinash Kaushik has an amazing in-depth look at definitions and models of multi-channel attribution. Even better, he makes what could be a dry, dull topic enlightening, engaging and entertaining (no small trick).

  14. […]
    I do believe that trying to assign proper credit to each specific digital channel is important and worthy of digital marketers’ time. I do want, however, to provide a sense of reality that no one will ever achieve this goal. Multi-channel attribution is incredibly complicated and, in my opinion, has multiple meanings. Avinash Kaushik has an excellent blog post where he discusses three different types of multi-channel attribution:

  15. […]
    Or consider what will happen when we finally bridge the current technological chasm that prevents us from properly tracking and attributing consumer behavior and conversion not just across marketing channels but across platforms (e.g. mobile vs desktop vs TV). Heck, imagine what will happen when marketing and advertising can read thoughts psuedo-telepathically?

  16. […]
    For example, consider what will happen when the television industry finally succumbs and reluctantly allows TV to shift from a “dumb” offline media format to a “smart” online media format that allows for robust advertising and marketing analytics and targeting. Or consider what will happen when we finally bridge the current technological chasm that prevents us from properly tracking and attributing consumer behavior and conversion not just across marketing channels but across platforms (e.g. mobile vs desktop vs TV). Heck, imagine what will happen when marketing and advertising can read thoughts psuedo-telepathically?

  17. […]
    To drive the point home, here’s an amazing quote from Jim Novo, from a conversation we had on Avinash’s blog (BTW – this post is a must-read for understanding the challenges of multi-channel attribution):

  18. […]
    If something will not add value to your customer or your business, invest your money elsewhere. Any badly done media or online entity (such as a blog, app or facebook page) can negatively affect your brand. I might be missing something but I have to make mention of Avinash Kaushick’s brilliant explanation of multi-channel attribution. He breaks it down into 3 types, all of which are referred to in conversations, and hence the confusion:

  19. […]
    Avinash Kaushik ( gave the keynote speech and was fabulous. Note that this keynote was free to anyone; so the take-away here is register for free and attend at least the keynotes. His big theme was on metrics – micro conversions, macro conversions, and economic value. Behind the amazing presentation was really the re-packaging of the theme of having a “marketing ladder” from free to value on your website. He gave us some good links – Data Driven Documents, Multi-channel Attribution, and focused a lot on intent marketing.

  20. […]
    The same thing applies to web analytics. In the early 2000, companies might have been looking at page views and bounce rates on their websites. Now, according to Avinash Kaushik, the focus is on multi-channel attribution ( In a sense, those who are in the analytics industry not only have to grasp on fundamentals such as reporting and dashboards but also have to learn how digital and traditional media channels fall into the holistic picture of businesses.

  21. […]
    Repeat after me: "MY USERS ARE GREATER THAN THE SUM OF THEIR KEYWORDS." Note: this is a helpful motto to keep in mind as we approach year two of "Keyword (not provided)." Instead, use multi-channel attribution measurement with a longer attribution period than you might normally employ (say, a 30-day window instead of just same-day or even seven-day attribution) in order to get a better sense of the customer's actions online.

  22. […]
    This is one of the many Multi-Channel Attribution problems we face, a popular topic in the measurement community. One of my favorites, Avinash Kaushik, articulated this, saying:

  23. […]
    Instead, use multi-channel attribution measurement with a longer attribution period than you might normally employ (say, a 30-day window instead of just same-day or even seven-day attribution) in order to get a better sense of the customer’s actions online.

  24. […]
    Préférez des mesures d’attribution multicanaux avec une période d’attribution longue (disons 30 jours) de manière à ce que les actions des visiteurs prennent plus de sens. Mike Pantoliano de Distilled a récemment parlé des modèles d’attribution à la MozCon en montrant comment commencer rapidement avec simplement Google Analytics. Vous découvrirez des phénomènes étonnants sur votre propre site lorsque vous appliquerez ceci :

  25. […]
    # Avinash Kaushiks Beiträgen zur Channel-Attributionsforschung,

  26. […]  
    I was flagged on this post because it has a quote from me that seems to support Rok’s thesis about the death of the funnel model and the related idea, “Direct Response Measurement is a Wet Dream”. The quote is from a comment I made on a post by Avinash where we were discussing the value of sequential attribution models:

  27. […]
    Von Google gibt es dazu eine sehr gute und viel zitierte Analyse (The New Multi-screen World, pdf Download). Einsteiger in das Thema sollten unbedingt auch “Multi-Channel Attribution: Definitions, Models and a Reality Check” von Avinash Kaushik lesen.

  28. […]
    Avinash wrote a brilliant post about this earlier this year. It's a pretty dense and heady read, but it's the best answer I've found for this very big topic.

  29. […]
    Avinash Kaushik has a great, long, in-depth post on how to build your own system to count points scored and assists contributed called Multi-Channel Attribution: Definitions, Models and a Reality Check that also tries to incorporate in-store purchases.

  30. […]
    In this excellent blog, Avinash Kaushik identifies three types of multi-channel attribution. The most common type and the most straightforward is multi-channel attribution across different digital channels, which is what most digital marketers refer to when they talk about attribution. Kaushik calls this MCA-ADC, and it involves determining the contributions of individual digital channels to a particular conversion. These digital channels can be in the form of organic search results, pay-per-click (PPC) ads in search engines, e-mail, referrals from an affiliate site, display ads, social media ads or YouTube videos. Through the use of cookies and other tags, it is possible to track the online behavior of individual users and their interactions with every digital channel up until a conversion. This is known as ‘path-to-conversion’ analysis.

  31. […]
    (Avinash Kaushik, ‘Multi-Channel Attribution: Definitions, Models and a Reality Check’, April 2012)

  32. […]
    In fact, it’s near impossible, and Avinash Kaushik clearly states why in his brilliant blog post Multi-Channel Attribution: Definitions, Models and a Reality Check (I highly recommend reading this!). He goes into subtle detail and highlights three core areas for attribution modelling:

  33. […] Multi-Channel Ascription: Connotations, Models plus a Reality Confirm ( […]

  34. […]
    Analyticsin attribuutiomallinnus toimii parhaiten ja tarkiten vain jos kaikki käynnit verkkosivustolla tapahtuvat samalla selaimella, poistamatta evästeitä. Toisin sanoen, seuraavat rajoitteet tulee tiedostaa ennen tulkintojen muodostamista:
    Analytics ei kykene seuraamaan käyttäjän koko polkua saumattomasti, mikäli käyttäjä vierailee sivustolla eri päätelaitteilla kuten tietokoneella, mobiililaitteella ja tabletilla.
    Analytics ei pysty seuraamaan puhelinsoittoja tai vierailuja kivijalkaliikkeeseen
    Laajempi kirjoitus aiheesta löytyy esimerkiksi Avinash Kaushikin blogista.

  35. […]
    Na kraju ovog teksta preporučujem “reality check” o ovoj oblasti iz pera Avinasha Kaushika, vodećeg stručnjaka iz oblasti Web analitike. U osnovi, kod nas je prvo potrebno da se kvalitetno nauči značenje i relano praćenje samih konverzija (tzv. last-click analiza), pa kada se to savlada, a tek onda da se stvari “dodatno komplikuju” sa “tamo nekim dodatnim doprinosima” efektima promocije, uključujući i napredne sisteme dodavanja rezultata offline prodajnih efekata.

  36. […]
    There are actually a few different types of Multi-Channel Attribution, ranging from the traditional ‘last click’ method through to a variety of multi-point models. Here is an overview of the most common types, and I encourage you to research each of them further as each one harbours complexities that warrant a deeper look:

  37. […]
    It’s been covered in far more detail elsewhere, but in a nutshell: attribution modeling attempts to solve the problem of which channel gets credit when a user touches multiple channels prior to converting. Many marketers simply throw up their hands and say the last touch gets all the credit – but then we have to live with the knowledge that some of our efforts are far more effective than we give them credit for.

  38. […]
    Chiudo questo lungo post, segnalando una lettura molto interessante di Avinash Kaushik al seguente link in cui si analizzano le sfide aperte dal multichannel attribution, sfide che vanno oltre il solo online ed investono anche l’analisi delle sinergie che si creano fra online-offline. Avremo modo di approfondire l’argomento prossimamente

  39. […] Multi-Channel Attribution: Definitions, Models and a Reality Check […]

  40. […]
    Multi-Channel Attribution: Definitions, Models and a Reality Check – Avinash Kaushik gives a high-level analysis of attribution modeling, defining problems with current models and offering suggestions to improve your own models.

  41. […]
    Az attribúciós modellek végső soron szintén a big data problémakörébe tartoznak. Alapvetően három fajta attribúciós probléma van: az online hirdetések offline hatása, a képernyők közötti megoszlás (tévé, asztali gép, tablet, mobil), ill. az egyes digitális csatornák közötti megoszlás. Leggyakrabban az utóbbival szoktunk foglalkozni.

  42. […]
    Avinash Kaushik’s post on Multi-Channel Attribution: Definitions, Models and a Reality Check

  43. […]
    Desweiteren wird eine große Aufgabe darin bestehen die einzelnen Marketing-Maßnahmen in Beziehung zueinander zu setzen und den Beitrag jeder einzelnen Maßnahme zum Ganzen zu erfassen (Multi-Channel-Tracking, Attribution- Modelling). Dieser Datenmengen Herr zu werden und die Interpretation wird dem Berufsbild des Marketing-Analysten neues Gewicht geben(Stichwort: Big Data). Bisher hat jede Marketing Disziplin ihr eigenes Süppchen gekocht. Zusammengefasst lässt sich das Marketing der Zukunft mit folgenden Attributen beschreiben:

  44. […]
    To quickly review, multi-channel attribution refers generally to the process of parsing out how, and to what extent, different consumer touch points influence consumers’ buying behaviors. However, different stakeholders have concentrated this definition differently. For clarity sake, we can distill these differences into three common focuses: (1) the impact of online communications on offline sales, (2) the consumer experience across multiple devices as it drives toward conversion, and (3) the consumer experience across multiple digital marketing channels as it drives toward conversion.[1]

  45. […]
    Como vemos los modelos de atribución nos permiten atribuir al online parte de las ventas del offline, sobretodo ahora con el nuevo Measurement Protocol de Analytics. Los modelos de atribución no solo nos permiten estudiar canales de venta. Avinash en su post sobre modelos de atribución habla además del estudio entre canales, de utilizarlos para el estudio de la atribución de distintos dispositivos o pantallas.

  46. […]
    Repeat after me: ”MY USERS ARE GREATER THAN THE SUM OF THEIR KEYWORDS.” Note: this is a helpful motto to keep in mind as we approach year two of “Keyword (not provided).” Instead, use multi-channel attribution measurement with a longer attribution period than you might normally employ (say, a 30-day window instead of just same-day or even seven-day attribution) in order to get a better sense of the customer’s actions online.

  47. […]
    Ca notă personală, eu cred că lucrurile trebuie privite şi din perspectiva unui funnel mai lung, în care consumatorul devine familiar cu brand-ul prin Social Media, Display sau SEO. Acesta prinde încredere şi se abonează la Newsletter, iar, ulterior, converteşte. E clar că, în acest context, celelalte canale au şi ele o contribuţie, iar un sistem de Multi Channel Attribution trebuie folosit.

  48. […]
    Muito provavelmente, não. A atribuição de valores por canal é um assunto muito delicado e cheio de detalhes mais complexos do que a distribuição por média aritmética permite cobrir. Se você estiver interessado em desbravar esta área, eu sugiro começar pelo artigo (em inglês) Multi-channel Attribution: Definitions, Models and a Reality Check, do Avinash Kaushik – no artigo, ele esclarece bastante sobre o assunto, especialmente sobre a complexidade envolvida em se fazer a atribuição/distribuição de méritos entre os canais de conversão.

  49. […]
    Some say they don’t produce immediate leads or sales and the ROI from multi-channel marketing is difficult to measure. However, just because it’s not as easy to attribute sales to these traditional channels doesn’t mean they are useless. Google, for example, has rolled out a tool to help with attribution modeling, its Universal Analytics, which recently was opened to anyone. And for years Google has been proving by example that it believes in traditional marketing channels.

  50. […]
    If your actual conversions occur offline – such as if a customer needs to make a phone call – close the gap between online and offline actions as much as possible. For example, consider a phone tracking service like that can attribute phone calls to your website. For more info, Avinash Kaushik goes into great depth on this: Multi-Channel Attribution: Definitions, Models and a Reality Check

  51. […]
    Ca notă personală, eu cred că lucrurile trebuie privite şi din perspectiva unui funnel mai lung, în care consumatorul devine familiar cu brand-ul prin Social Media, Display sau SEO. Acesta prinde încredere şi se abonează la Newsletter, iar, ulterior, converteşte. E clar că, în acest context, celelalte canale au şi ele o contribuţie, iar un sistem de Multi Channel Attribution trebuie folosit.

  52. […]
    The source of lead is where you can draw easier connections to your SEO activity. You will also find that the source matches up quite nicely with an assisted conversion. This is really a more in-depth discussion on multi-channel attribution than we have time and space for here, but Google Analytics can handle that, too.

  53. […]
    Start with the “Multi-Channel Attribution: Definitions, Models and a Reality Check” posting. It provides an introduction to the problem, an overview of the complexities involved and several modeling approaches. This post also contains links to several earlier posts that describe a number of techniques to help get you started. (As an added bonus, look for pictures of the type of vacation that you’ll be able to afford if you get really good at this stuff!)

  54. […]
    What I felt today was different, though. It came when I was looking for some info on a piece of software I use (Google Analytics). I started at the website of a well-known expert in the field, Avinash Kaushik at his The Good, The Bad, and The Ugly piece. That article led me to his Definitions, Models, and a Reality Check piece. That article led me to his Tracking the Online Impact of your Offline Campaigns piece, and that article led me to David Hughes’ website because he coined the term “non-line” that Avinash uses (it means marketing efforts that exist both on- and offline, like the color of a logo).

  55. […]
    Moving beyond the last-click attribution model allows marketers to look at the full funnel of touch points that contributes to a sale or conversion. It’s called multi-channel attribution. This more advanced practice allows marketers to distribute credit for a conversion among all digital touch points, and in turn gain new insights, optimize campaigns, and get a more accurate reading on ROI.

  56. […]
    Multi-Channel Attribution: Definitions, Models and a Reality Check

  57. […]  
    Conversions/Revenue from Organic Traffic (Non-Branded and Branded) – This is also obfuscated by Not Provided. Beyond that, almost all conversion funnels should be taking multi-channel attribution into account and without that, you’re left with last click conversion data, which isn’t an accurate representation of your actual conversion funnel. If you need some convincing on the value of multi-channel attribution, go read some of the awesomeness that is Avinash.

  58. […]
    And they do have a lot of great stuff coming up in Google Universal (GU from here on). The vision is obvious in its simplicity – to break down the barriers that exist between silos of information whether online or not. So today, when we look at the attribution models that exist there are a lot of things missing in the model (a lot of these are straight from Avinash’s posts – who works for Google as well, particularly his post here):

  59. […]
    Attribution seems to be a problem that is growing in complexity in digital marketing. For those not familiar with it, it’s the process of figuring out what event caused a desired outcome. For example, did a certain page on a website cause a customer to purchase a service. While that specific attribution problem can now easily be solved, there are much larger problems like multi-channel attribution, and multi-device attribution. Now the questions are more like did the post a person read on Google+ with their mobile phone effect their decision to buy when they looked a website on their laptop 3 hours later. There’s already been some good discussions on this, so that’s not the purpose of what I’m talking about.

  60. […]
    Yakın bir gelecekte birbiri arasında bir kullanım tecrübesi akışı sağlayan cihazların kullanımlarının artışı ile artık reklamın önce hangi kanalda görüldüğü veya tıklandığı değil, hangi cihazda bu tecrübenin yaşandığı da ciddi oranda önem kazanacak. Burada kısaca değindiğim ekranlar arası akış raporlaması da Avinash Kaushik’in belirttiği gibi aşağıdaki karmaşıklıkta bir tüketici davranışının anlamlı parçalara bölünerek ayrıştırılma analizini karşımıza çıkaracak.

  61. […]
    Thankfully, most of the top analytics platforms offer some multichannel attribution reports for you to leverage. Before you begin implementing multichannel attribution, however, you should be aware of the technical and strategic challenges this entails (Avinash Kaushik's post does a great job of explaining some of these difficulties and how to approach them). But don't let the difficulty deter you! The multichannel attribution approach allows for a much more accurate understanding of how the entire marketing mix contributes to results.

  62. […]
    The technical term for solving this difficult puzzle is called "sales source attribution". Because our shoppers are invisible and we are a ROBO industry (Research Online, Buy Offline), I felt the clearest picture would come from a delivery survey. Survey at Delivery: A Road Map to Your Store: Survey at Delivery: A Road Map to Your Store Its easy to create and run and you'll beable to "see" those damn autotrader shoppers that are so hard to see (until they buy). Lastly, I'll leave you with some homework from the grand master, Avinash: Multi-Channel Attribution: Definitions, Models and a Reality Check

  63. […]
    Plenty of talk recently about multi-channel attribution and I fell back on this great (but very long) post from Avinash Kaushik. The opening line ‘To guarantee success, spend 95% of your time defining the problem and 5% of the time solving it’ always resonates, not just digitally.

  64. […]
    This information is extremely valuable for marketers so that they can determine which channels to invest in more actively. For a more advanced understanding of the different types of multi-channel attribution models, this blog post at Occam’s Razor does a masterful job. For our purposes, we’ll focus on the issue as it pertains to paid search.

  65. […]
    Track the person, not the device. Other people (like Avinash Kaushik and Craig Bradford) can explain this much better than me, but the short version is: Stop tracking each session as if it’s a different user. Instead, track people throughout their journey from start to finish – irrespective of device.

  66. […]
    Step two, once they know you exist, is where that 3D spatial map may become relevant as a way to visually chart the choppy navigational waters of interest/consideration. This is certainly true in the long B2B buying cycle, and now with cross-devices, multiple channels, and the complexity of modern marketing and the utter complexity of marketing attribution, it’s true of the B2C market as well.

  67. […] Track the person, not the device. Other people (like Avinash Kaushik and Craig Bradford) can explain this much better than me, but the short version is: Stop tracking each session as if it’s a different user. Instead, track people throughout their journey from start to finish – irrespective of device.

  68. […]
    Attribution administration helps break down silos, incorporating all online and offline execution to give a comprehensive picture of execution and empowering marketers to unmistakably and precisely recognize which channels really prompted or impacted a transformation. Here is a super authoritative post on how Multi channel attribution works.

  69. […]
    Determinar el Modelo de Atribución más adecuado para tu empresa puede llevarte tiempo, no desesperes, como en tantos otros aspectos del mundo digital el sistema prueba-error te ayudará a determinar qué Modelo se ajusta mejor a tu negocio. Recuerda que: “La energía y la perseverancia conquistan todas las cosas” (Benjamin Franklin). ¿Quieres saber más sobre Modelos de Atribución Multicanal?, no te pierdas el post de Avinash K. “Multi-Channel Attribution: Definitions, Models and a Reality Check”.

  70. […]
    Akhirnya saya Googling dan masuklah ke artikelnya Avinash Kaushik sang digital analytics guru Multi-Channel Attribution: Definitions, Models and a Reality Check. Avinash Kaushik memang keren. Gaya penulisannya ringan, padahal topiknya sangat berat. Dari sana, barulah saya memahami duduk permasalahan sebenarnya.

  71. […]
    Multi-Channel Attribution : Definitions, Models and a … – Apr 02, 2012 · What multi-channel attribution models deliver value when marketing across devices (TV, phones, tablets) & channels (search, display, social)?…

  72. […]
    As your understanding of attribution modeling improves, you will start to ask questions that can be resolved with more detailed modeling. The MT model we discussed above is a version of Avinash Kaushik’s Across Digital Channels models:

  73. […]
    Start small with a clearly defined framework and set of metrics for judging success. I’m not going to tackle attribution modeling in this post, but it’s definitely something you need to be aware of, and for now, you can get more information on it here — Multi-Channel Attribution: Definitions, Models and a Reality Check.

  74. […]
    And they do have a lot of great stuff coming up in Google Universal (GU from here on). The vision is obvious in its simplicity – to break down the barriers that exist between silos of information whether online or not. So today, when we look at the attribution models that exist there are a lot of things missing in the model (a lot of these are straight from Avinash's posts – who works for Google as well, particularly his post here):

  75. […]
    Desweiteren wird eine große Aufgabe darin bestehen die einzelnen Marketing-Maßnahmen in Beziehung zueinander zu setzen und den Beitrag jeder einzelnen Maßnahme zum Ganzen zu erfassen (Multi-Channel-Tracking, Attribution- Modelling). Dieser Datenmengen Herr zu werden und die Interpretation wird dem Berufsbild des Marketing-Analysten neues Gewicht geben(Stichwort: Big Data). Bisher hat jede Marketing Disziplin ihr eigenes Süppchen gekocht. Zusammengefasst lässt sich das Marketing der Zukunft mit folgenden Attributen beschreiben:

  76. […]
    Viele Analyse-Tools in Websites, Landingpages und eCommerce Shops arbeiten nach dem “Last-Click-Attribution“-Modell, dh. ein Kauf wird dem letzten Klick zugeordnet. Dass diese Messmethode heute nicht mehr der normalen Customer Journey entspricht, zeigen viele unterschiedliche Studien. Entsprechend messen immer häufiger Werbetreibende nach dem Multi-Attribution-Modell, dh. es werden die verschiedenen Touchpoints innerhalb der Customer Journey bis zu einem Kauf, bzw. einer Conversion, ermittelt. View Tags unterstützen die Analyse nach dem Multi-Attribution-Modell.

  77. […]
    First and last interaction attribution are relatively simple to setup and track. They also provide quick insights for “pulse” readings. Additionally, they can be highly useful when tracking mobile activity given that many mobile networks do not track impressions but only clicks.

  78. […]
    Multi-channel attribution is the holy grail for any analyst, or at least that’s what we’ve all been led to believe. In the simplest terms, multi-channel attribution across marketing channels is assigning a value to each touch point along the way to a conversion. But it’s easier said than done.

  79. […]
    So, you are interested in learning how digital marketing channels such as Display and YouTube contributed to a conversion (or multiple conversions). In other words, this is what Avinash Kaushik calls MCA-ADC: Multi-Channel Attribution Across Digital Channels. You’ve already seen this other Last-Click blog article and as a result you enabled your Google Analytics account to tie display impressions to your conversions.

  80. […]
    Attribution modeling is tough (Go read this post from Avinash Kaushik), and most companies are still trying to figure this out. But before you completely discount a channel, dig deeper into your analytics. Look at both first-click conversions and last-click conversions, look at assisted conversions, and most importantly, evaluate conversion paths.

  81. […]
    Vanzelfsprekend zijn er een aantal manieren om bijvoorbeeld ook de offline wereld goed door te meten maar dit is nooit alomvattend en we zullen voorlopig nog niet in staat zijn om elk touchpoint volledig op de juiste waarde te schatten. Avinash Kaushik heeft hier, in zijn blog Multichannel Attribution Models, een zeer goede post over geschreven (leestip!).

  82. […]
    The real impact for long-cycle products like cars and home loans may never be known with precision. But methods continue to improve. It is now possible to track individual ad exposures in digital channels directly to sales in ways that were impossible five years ago. If real value can be shown to a CMO, money will follow.

  83. […]
    We must keep in mind that looking at attribution solely from a digital perspective leaves certain gaps in the data, particularly when a customer interacts with us across multiple devices or when a customer begins their buying journey online but completes the conversion (purchase) in the store. You can read more about those challenges here at Multi-Channel Attribution: Definitions, Models, and a Reality Check by Avinash Kaushik.

  84. […]
    Once you have this information, set up a dashboard in Google Analytics or whatever platform you use to monitor key metrics related to business performance. At a minimum, your dashboard should contain: visits (sessions) time on page engagement with each social platform — shares, likes, comments, RT page value from multichannel attribution modeling

  85. […]
    Esse artigo é uma tradução do texto original do Avinash Kaushik, evangelizador de marketing digital da Google. Já dizia um sábio: "Para garantir sucesso, gaste 95% do seu tempo definindo o problema e 5% do tempo resolvendo-o." Eu acredito veementemente nessa frase. Na minha vida, passo uma quantidade extraordinária de tempo entendendo o problema e tentando defini-lo claramente. Como se fosse mágica, eu descobro que dessa maneira é muito mais fácil encontrar a melhor solução (ou perceber que nenhuma existe!).

  86. […]
    Jest również autorem dwóch najlepiej sprzedających się książek Web Analytics 2.0 i Web Analytics: An Hour a Day. Te książki to obowiązkowe lektury, dla kogoś kto chce się zająć analityką internetowa. Na blogu znajdziesz wiele informacji o korzystaniu z segmentacji i wielokanałowej atrybucji. Warto przeczytać Multi-Channel Attribution: Definitions, Models and a Reality Check

  87. […]
    He therefore recommends heuristics driven models such as ‘Last Click’, ‘First Click’, ‘Linear Interaction’, and ‘Time Decay’ etc. to determine attribution. ( This of course stated in 2012, and since then a lot has changed and much more data has flown down the digital highway.
    This may be a good time to take you through the heuristics that are widely used in digital attribution. Green shows the extent of attribution.

  88. […]
    Helaas zijn er weinig dingen zo ingewikkeld als attributie binnen analytics. Maar laat ons toch een poging doen om de verschillende attributiemodellen wat beter te begrijpen. In zijn artikel definieert Avinash Kaushik 3 grote uitdagingen voor attributie binnen online marketing:

Add your Perspective